--- title: "Agents That Run Outbound While You Sleep" episode: 92 podcast: "The LeanScale Podcast" publisher: "LeanScale" guest: "Mica" guest_title: "Founder & CEO" date_published: 2026-07-13 date_modified: 2026-07-22 duration: 00:51:33 word_count: 9856 topics: ["outbound-sales", "ai-in-gtm", "gtm-strategy", "demand-generation", "sales-leadership"] canonical_url: https://leanscale-knowledge-hub.netlify.app/podcast/mica-ample-market-outbound-agents/ source: "LeanScale Podcast Knowledge Hub — https://leanscale-knowledge-hub.netlify.app" license: "Free to quote and cite with attribution to The LeanScale Podcast." --- # Agents That Run Outbound While You Sleep _Mica, founder & CEO of Amplemarket, on agentic GTM — building outbound agents that research, personalize, and run while you sleep_ **Episode 92 · The LeanScale Podcast** Mica, Founder & CEO (Ample Market) · Hosted by Anthony Enrico Published July 13, 2026 · Updated July 22, 2026 · 00:51:33 Canonical: https://leanscale-knowledge-hub.netlify.app/podcast/mica-ample-market-outbound-agents/ **Topics:** Outbound & Sales Development · AI in GTM · GTM Strategy · Demand Generation · Sales Leadership ## Executive summary For a year the loudest story in GTM has been that AI plus a couple of engineers quietly shrinks the sales org. Mica — founder and CEO of Amplemarket, the consolidated go-to-market platform used by thousands of sales and GTM teams globally — spends this episode arguing the opposite, and then proving it live. In a demo-heavy conversation with LeanScale co-founder Anthony Enrico, he wires Claude to Amplemarket's new MCP and agent layer and walks through agents doing real GTM work: pulling a podcast's guest list and enriching it in one prompt, mapping the VP-of-sales at a conference's top 20 sponsors from a logo grid, parsing a screenshot of Twitter followers into a sequenced outbound list, and triggering bulk SDR connection requests from a single Telegram message. His organizing idea is that agentic GTM is built from a handful of interchangeable building blocks — 'Lego pieces': a very smart intern (the LLM), web search, a data platform, and a workflows agent that can take or suggest actions. The hard part isn't capability, it's knowing where to start, which is why he pushes operators to whiteboard the recurring work a smart new hire could own. He layers on an adoption ladder most GTM teams haven't climbed: use skills and you're a top 1% AI user; use scheduled tasks and you're top 0.1%. Skills are precise instruction sets that sharpen an LLM's output; scheduled tasks are routines that wake an agent at, say, 3:50 AM to do the work before anyone logs in. The most important pattern he demonstrates is the two-agent handoff: a research-and-enrichment agent running in Claude passes its output to an engagement agent living inside Amplemarket, which already owns the channels of communication and sits on the CRM's context. That second agent is where personalization stops meaning first-name swaps. Because it knows whether you're already connected, whether you emailed or cold-called the prospect before, and what the CRM says about the account, it decides not just the copy of each step but the channel each step starts on. Scheduled agents extend the same idea across time — a 3:50 AM Product Hunt monitor qualifies overnight launches for B2B fit and routes them to the right rep; a Slack watcher turns inbound-lead activity into suggested multi-threading into C-level. On the 'is outbound dead?' debate, Mica is blunt: templated, non-contextual, mass-volume outbound is dead — the 'hey {first name}, I saw you're at {company}' motion in Outreach and Salesloft — but contextual, signal-based outbound is thriving. His proof point is Amplemarket's own numbers since January 2026: opportunities generated split roughly 40% email, 40% calling, 20% social, with the company driving millions in pipeline every month running its own product. Anthony reframes it as 'lazy outbound is dead' — people connecting to solve real problems for each other never goes away. The closing arc is the headcount question. Both reject the layoff thesis on market logic: if a tool makes a team dramatically faster, competitors force you to reinvest that speed, not bank it — so productivity gains grow BDR and AE teams rather than gutting them, and the real risk is a widening productivity gap between operators who adopt and those who don't. Anthony offers LeanScale as a live case: AI-driven diagnostics, transcript-fed project management, and account-health agents raised delivery quality and margin without firing anyone, because the market simply reset expectations higher. Who should listen: founders, RevOps and GTM operators, and SDR/sales leaders who want a concrete blueprint for agentic outbound — and a grounded, optimistic read on what the next 12 months of AI in GTM actually looks like. ## Key takeaways 1. **Design agentic GTM from four building blocks, not from a product** — Mica frames agentic GTM as a set of interchangeable 'Lego pieces': a very smart intern (the LLM), web search, a data platform, and a workflows agent that can take or suggest actions on your behalf. Any GTM play you can imagine is a different combination of those blocks. _Why it matters:_ Stop shopping for a magic tool and start whiteboarding the recurring work a smart intern could own. Decide which blocks you have and which you're missing, then compose plays from them rather than waiting for a packaged 'set.' _For:_ RevOps Leaders, Founders, Sales Leaders 2. **Skills make you top 1%; scheduled tasks make you top 0.1%** — A skill is a precise instruction set that sharpens an LLM's output and, for Amplemarket customers, installs in one click. A scheduled task is a routine that runs the agent on a cadence — daily, weekdays, or a specific time like 3:50 AM. Even in AI-pilled San Francisco, Mica finds most GTM people using neither. _Why it matters:_ Adopting skills and scheduled tasks is the cheapest way to leapfrog most of the market. The productivity gap opening up in GTM is largely a function of who has climbed this ladder and who hasn't. _For:_ RevOps Leaders, Sales Leaders, Marketing Leaders 3. **An MCP turns your LLM into an operator of your GTM stack** — An MCP connects your platforms into your LLM so Claude or ChatGPT can call your data and engagement layer inside a chat. Because Amplemarket already owns the channels, sits on the CRM, and holds a strong business-data engine, one prompt can enrich, decide, and act instead of just returning text. _Why it matters:_ The leverage isn't the model alone — it's the model wired to a data-and-engagement platform via MCP. Invest in the connective tissue so your agents can take actions, not just draft them. _For:_ RevOps Leaders, Founders, Revenue Executives 4. **Chain one prompt from raw source to enriched, sequenced outbound** — The demos collapse a multi-tool workflow into a single prompt: find a podcast's last 10 guests, infer who they are, enrich emails and LinkedIn, add an insight and icebreaker, and build a list — or take a conference page, convert sponsor logos to companies, find their VPs of sales, and enrich them. Even a screenshot of Twitter followers becomes a validated list. _Why it matters:_ Friction is what kills good plays — the list-find-export-enrich-sequence chain has enough steps that reps skip it. Removing that friction 'rips the ceiling off' of how much high-quality outbound a team can actually run. _For:_ Sales Leaders, RevOps Leaders, Marketing Leaders 5. **The two-agent pattern: a research agent hands work to an engagement agent** — A Claude agent does the parsing, research, and enrichment, then hands off to a second agent living inside Amplemarket that has its own instructions for how to act. One agent is plugged into the web; the other is harnessed to your communication channels and CRM. They communicate, and every workflow is a distinct agent. _Why it matters:_ Separate 'find and understand' from 'engage and act.' The research agent maximizes context from the open web; the engagement agent maximizes context from your owned channels and CRM — together they produce action, not just insight. _For:_ RevOps Leaders, Sales Leaders, Founders 6. **Real personalization decides the channel, not just the copy** — This isn't a seven-step sequence with a swapped first name. Because the engagement agent knows whether you're already connected, whether you emailed or cold-called before, and what the CRM says, it decides for each prospect whether to start on LinkedIn, email, a call, or a text — and writes each step bespoke to that person's 'who' and 'why.' _Why it matters:_ Per-prospect personalization means the entry channel and sequence are generated from scratch per person, grounded in prior interactions and CRM history. Generic multi-step templates are exactly the outbound that's now dead. _For:_ Sales Leaders, RevOps Leaders 7. **Scheduled agents do the work while you sleep** — Mica's Product Hunt agent wakes on weekdays at 3:50 AM, scrapes the last 24 hours of launches, qualifies each for B2B fit, extracts insights, and routes qualified ones to the right rep inside Amplemarket. Another monitors revenue podcasts weekly; another watches a Slack 'bot inbound' channel for demo requests. _Why it matters:_ The highest-leverage agents run unattended on a cadence, so the work is done before the team logs in. Point an agent at any webpage that acts as a ledger of new entries you'd normally have to react to. _For:_ RevOps Leaders, Sales Leaders, Founders 8. **Trigger agents from where you already work — Telegram and Slack** — After an in-person meeting, Mica messages his 'Hermes' agent on Telegram ('I just met Anthony, send him a connection request'), which hands off to an Amplemarket workflow agent. One Telegram message fired connection requests to an entire SDR team he'd just met. Slack messages can likewise trigger monitoring and action. _Why it matters:_ Agents shouldn't require you to be in the tool. Wire triggers into the channels you already live in so a single message on the go kicks off enrichment, sequencing, and outreach automatically. _For:_ Sales Leaders, RevOps Leaders 9. **Reps won't leave frontier LLMs for vendor-native agents** — Mica doubts the model where vendors bury agents inside their own platforms and pull reps out of Claude or ChatGPT. Operators will always want the most intelligent model, the newest connectors, and advanced constructs like loops and goals — then plug those into their existing flows. _Why it matters:_ Build for a world where reps operate from the best available LLM and reach into your platform via MCP — not one where you try to keep them inside your UI. The intelligence layer wants to be the frontier model. _For:_ Founders, Revenue Executives, RevOps Leaders 10. **Use the 'smart intern' heuristic to design agents — and just ask AI how** — The design question is simple: if you hired a very smart intern tomorrow, what recurring work would you hand them? Build your agentic flows around that answer. And you don't need to write the instructions yourself — tell Claude what you want and it will interview you and write the scheduled task for you. _Why it matters:_ Agent design is a delegation exercise, not an engineering one. Naming the intern's job list is the real work; the model handles the setup, so the barrier to entry is far lower than it looks. _For:_ RevOps Leaders, Founders, Sales Leaders 11. **Outbound isn't dead — templated outbound is** — The dead motion is non-contextual, mass-volume, first-name-swap outbound of the kind run in Outreach and Salesloft. Contextual, signal-based outbound works: Amplemarket's own opportunities since January 2026 split roughly 40% email, 40% calling, 20% social, and the company drives millions in pipeline monthly running its own product. _Why it matters:_ Kill the templated sends, not the channel. Email works when it's context-rich and reaches the right person at the right time; social converts but is volume-capped; multi-channel, signal-based outbound remains a durable pipeline engine. _For:_ Sales Leaders, Revenue Executives, Marketing Leaders 12. **AI is a velocity multiplier — the market grows teams, it doesn't cut them** — Both reject the layoff thesis on market logic: if AI lets you go faster, competitors force you to reinvest that speed rather than bank it, so you don't fire people to move at the old pace. Mica's fastest-growing team is his SDR team; he sees customers doubling and tripling BDR and AE orgs. The real risk is a widening productivity gap between adopters and everyone else. _Why it matters:_ Plan for reinvestment, not reduction. Spend AI gains on velocity and quality — as LeanScale did internally without firing anyone — because the market resets expectations higher, and the operators who adopt fastest leave the rest in the dust. _For:_ Founders, Revenue Executives, Sales Leaders ## Frameworks ### The Lego Pieces (Building Blocks of Agentic GTM) (02:37) **Definition:** Agentic GTM is composed from a few interchangeable building blocks: a very smart intern (the LLM), web search, a data platform, and a workflows agent that can take or suggest actions on your behalf. Like Lego, the pieces recombine into any play — but a pile of loose bricks is useless without a design, which is why Mica pushes operators to whiteboard the recurring work a smart intern could own before reaching for a tool. ### The AI Adoption Tier Ladder (Skills → Scheduled Tasks) (30:10) **Definition:** A maturity ladder for AI in GTM: using skills (precise, installable instruction sets) puts you in the top 1% of users; using scheduled tasks (routines that run agents on a cadence) puts you in the top 0.1%. Most GTM teams, even at AI-first companies, sit below the first rung. Climbing it is the cheapest way to leapfrog the market and is the main driver of the productivity gap opening up between operators. ### The Two-Agent Pattern (Research → Engagement) (19:43) **Definition:** A research-and-enrichment agent (running in the LLM, plugged into the web) hands its output to an engagement agent living inside the GTM platform (harnessed to owned channels and CRM), which has its own instructions for how to act. Separating 'find and understand' from 'engage and act' lets each agent maximize a different context source — the open web versus your CRM and channels — and produce action instead of just insight. Every workflow is its own distinct agent. ### Truly Per-Prospect Personalization (Who → Why → Channel) (24:52) **Definition:** Personalization that generates a bespoke sequence per person: the agent knows whether you're connected, whether you emailed or cold-called before, and what the CRM says, and decides both the copy of each step and the channel each step starts on. It propagates a specific 'who' with a specific 'why' (AE/SDR notes, closed-lost reasons, prior conversations), so the entry point and content are built from scratch per prospect — the opposite of a templated multi-step sequence with a swapped first name. ### The Smart-Intern Heuristic (37:24) **Definition:** To design agentic flows, ask what recurring work you would hand a very smart intern you just hired — then build agents around that answer, letting the LLM interview you and write the instructions. It reframes agent design as delegation rather than engineering: naming the intern's job list is the real work, and the model handles the setup, so the barrier to entry is far lower than it appears. ### The 40/40/20 Channel Mix (41:20) **Definition:** Across Amplemarket customers since January 2026, opportunities generated by channel split roughly 40% email, 40% calling, and 20% social — evidence that multi-channel, contextual outbound still works. Email works when it's context-rich and well-targeted; calling remains a heavy contributor; social converts well but is capped by connection-request volume. The mix disproves the 'email is dead / outbound is dead' narrative for contextual motions. ### Productivity Inequality → Market Equilibrium (47:17) **Definition:** AI creates a widening productivity gap between adopters and non-adopters, but market pressure keeps teams employed: if you can go faster, competitors force you to reinvest that speed rather than fire people to move at the old pace. The narrative of a GTM 'job apocalypse' is broken — reinvestment, not reduction, is the rational response, which is why Mica's SDR team is his fastest-growing and customers are doubling BDR/AE orgs while quality expectations reset higher. ## Quotes _Speakers inferred from an undiarized transcript — verify before attributing._ > "To me, these are like Lego pieces — different colors, and we can just build anything." > > — Mica, The LeanScale Podcast Ep. 92 (03:12) > "If you want to become a top 1% user of AI in GTM, just understanding what a skill is and leveraging skills can actually get you there." > > — Mica, The LeanScale Podcast Ep. 92 (07:16) > "If you're using skills, you may be top 1% AI users. If you're using scheduled tasks, then they are top 0.1% AI users." > > — Mica, The LeanScale Podcast Ep. 92 (30:10) > "You switch off your display and your Claude agent will wake up at 3:58 a.m. and it's going to execute those instructions for you." > > — Mica, The LeanScale Podcast Ep. 92 (30:57) > "To be able to do the whole thing from Claude with the backend of Ample Market giving you all the data you need, that just really unlocks the ceiling of what your outbound efforts could be." > > — Anthony Enrico, The LeanScale Podcast Ep. 92 (15:16) > "This isn't saying we're going to run a sequence that has seven steps and personalize it by changing their name. No, no, no. Each person is getting their sequence started from scratch." > > — Anthony Enrico, The LeanScale Podcast Ep. 92 (24:14) > "Not only is the content of every single step truly bespoke and unique, but even where you're taking actions is unique." > > — Mica, The LeanScale Podcast Ep. 92 (25:22) > "How can you get Claude and ChatGPT to work for you without you being there?" > > — Mica, The LeanScale Podcast Ep. 92 (36:43) > "If you had a very smart intern that you just hired, what would you ask them to do? I would actually operate that way with your agentic flows." > > — Mica, The LeanScale Podcast Ep. 92 (37:24) > "The part of outbound that is dead is the templated outbound — 'Hey, first name, I noticed that you are at company name, let's do a call.' That's dead." > > — Mica, The LeanScale Podcast Ep. 92 (40:47) > "Outbound is not dead at all. If you take a look at opportunities generated by channel, it was 40, 40, 20 — 40% email, 40% calling, and 20% social." > > — Mica, The LeanScale Podcast Ep. 92 (41:20) > "We run Ample Market to do sales for Ample Market. My entire sales team, what they use is just Ample Market — and we generate millions of dollars in pipeline every single month." > > — Mica, The LeanScale Podcast Ep. 92 (42:37) > "Lazy outbound is dead, but really good signal-based outbound — people connecting to find the right solutions for each other — probably is never going to go away." > > — Anthony Enrico, The LeanScale Podcast Ep. 92 (43:18) > "If you can go faster, you're not going to fire people to go at the same speed you were going before. Because all of a sudden, within the market, your competitors will go faster." > > — Mica, The LeanScale Podcast Ep. 92 (47:17) > "I'm not worried at all for sales people. My fastest growing team has been my SDR team." > > — Mica, The LeanScale Podcast Ep. 92 (47:48) > "We didn't fire anyone. We just increased the quality that we're delivering and we're passing that on to the customer." > > — Anthony Enrico, The LeanScale Podcast Ep. 92 (49:04) > "Now that I can do unlimited work, there's more work to be done. It just completely ripped the ceiling off of what was on top of work." > > — Anthony Enrico, The LeanScale Podcast Ep. 92 (49:46) ## Practical advice by role ### Founders - Whiteboard the recurring GTM work a very smart intern could own, then compose agents from four building blocks: an LLM, web search, a data platform, and a workflows agent that can act. - Don't try to trap reps inside your product's UI — build for a world where they operate from the best frontier LLM and reach your data and engagement layer through an MCP. - Reinvest AI productivity into velocity and quality, not layoffs; the market resets expectations higher, so speed compounds into more pipeline and more hires, not fewer. ### RevOps Leaders - Climb the adoption ladder: install skills for precise, repeatable output, then move to scheduled tasks so agents run the work on a cadence before the team logs in. - Wire an MCP between your LLM and your data/engagement platform so a single prompt can enrich, decide, and act — not just draft text. - Adopt the two-agent pattern: a research/enrichment agent gathers web context and hands off to an engagement agent grounded in your CRM and owned channels. - Point a scheduled agent at any webpage that behaves like a ledger of new entries (launches, sign-ups, event lists) that would normally trigger a manual reach-out. ### Sales Leaders - Kill templated, first-name-swap sequences; make personalization decide the entry channel (LinkedIn, email, call, or text) per prospect based on prior interactions and CRM history. - Keep a human-in-the-loop checkpoint for high-stakes outreach and let low-stakes actions (like connection requests) run fully autonomously. - Trigger agents from where reps already work — a Telegram or Slack message after a meeting can fire connection requests and intro sequences automatically. ### Marketing Leaders - Treat launch and event signals as pipeline fuel: monitor Product Hunt, conference sponsor lists, and relevant podcasts with scheduled agents, then route qualified accounts to the right rep. - Feed real context (account history, closed-lost reasons, prior touches) into the engagement layer so outbound reads as human and relevant, not AI-generated. ### Revenue Executives - Judge outbound by contextual quality, not volume: the 40/40/20 email/calling/social mix shows multi-channel still generates opportunities when it's signal-based. - Frame AI investment as a velocity lever against competitive pressure — plan to grow BDR and AE capacity as productivity rises, and watch the widening adopter/non-adopter gap. ## AI takeaways **Thesis:** Agentic GTM is a craft of composition: wire a frontier LLM (Claude or ChatGPT) to a data-and-engagement platform through an MCP, teach it with skills, and let scheduled tasks run it while you sleep. Done this way, AI is a velocity multiplier that grows GTM teams and widens the gap between adopters and everyone else — not a headcount cut. - **Compose from building blocks** — Any agentic GTM play is a combination of four Lego pieces — a smart LLM, web search, a data platform, and a workflows agent that can act. The scarce skill is designing the play, not the capability. - **Climb the adoption ladder** — Skills (precise, installable instruction sets) put you in the top 1%; scheduled tasks (agents that run on a cadence) put you in the top 0.1%. Most GTM teams sit below the first rung. - **MCP makes the model an operator** — An MCP lets the LLM call your data and engagement layer inside a chat, so one prompt can enrich, decide, and take action — not just return text. The leverage is the model wired to the platform. - **Two agents beat one** — A research/enrichment agent on the open web hands off to an engagement agent grounded in your CRM and owned channels, which decides both the copy and the channel of each step per prospect. - **Reinvest, don't reduce** — AI productivity gains get spent on velocity because competitors force the pace — so SDR/BDR/AE teams grow, and the real risk is a widening productivity gap between operators who adopt and those who don't. - **Contextual outbound is alive** — Templated, first-name-swap outbound is dead; signal-based, multi-channel outbound (40/40/20 email/calling/social) still drives millions in pipeline when it reaches the right person at the right time. **Agent & automation ideas** - A 3:50 AM scheduled agent that scrapes the last 24 hours of Product Hunt launches, qualifies each for B2B fit, extracts launch insights, and routes qualified ones to the right account owner. - A one-prompt list-builder that pulls a podcast's or event's guest list, infers identities, enriches emails and LinkedIn, and adds a per-person insight, icebreaker, and connection message. - A conference-sponsor mapper that converts a sponsor logo grid into companies, finds their VPs of sales/revenue and GTM leaders, and enriches them into a target list. - A screenshot-to-sequence agent that parses an image (badge photo, follower list) into validated contacts and pushes them into a personalized sequence. - A Slack watcher that monitors an inbound-lead channel for demo requests and hands them to an engagement agent for immediate, context-rich outreach and C-level multi-threading. - A Telegram-triggered 'follow-up' agent (Hermes-style) that turns a post-meeting message into connection requests and intro sequences for one person or a whole team. - A weekly revenue-podcast monitor that tracks new episodes across shows and routes relevant guests/accounts to the right rep. ## Operations takeaways ### Revenue operations - **Build the connective tissue.** The leverage is an LLM wired to a data-and-engagement platform via MCP; invest there so agents can act, not just draft. - **Adopt skills, then scheduled tasks.** Skills sharpen output and install in one click; scheduled tasks run agents on a cadence so work happens before the team logs in. - **Separate research from engagement.** Run a two-agent pattern — a web-facing research agent hands enriched context to a CRM-and-channel-grounded engagement agent. - **Ground actions in CRM context.** Feed AE/SDR notes, closed-lost reasons, and prior touches into the engagement agent so it propagates a specific 'who' with a specific 'why.' - **Design agents like delegation.** Ask what a smart intern would own, then let the LLM interview you and write the scheduled-task instructions — the barrier is lower than it looks. - **Reinvest productivity.** Spend AI gains on velocity and quality; the market resets expectations higher, so throughput grows the team rather than shrinking it. ### Pipeline & marketing ops - **Remove workflow friction.** Collapsing find → enrich → sequence into one prompt is what lets reps actually run high-quality plays instead of skipping them. - **Personalize the channel, not the name.** Decide per prospect whether to open on LinkedIn, email, call, or text based on connection status and interaction history. - **Turn signals into pipeline.** Launches, event sponsors, follower lists, and podcast guests are raw signals a scheduled agent can convert into routed, enriched outbound. - **Trust the 40/40/20 mix.** Email and calling each drive ~40% of opportunities and social ~20% when outbound is contextual — the channel isn't dead, the template is. - **Keep a human checkpoint where it matters.** Let low-stakes actions run autonomously and preview high-stakes writing before it sends, so autonomy scales without losing judgment. ## Metrics mentioned | Value | Metric | Context | | --- | --- | --- | | 40% email · 40% calling · 20% social | Opportunity channel mix | Across Amplemarket customers since January 2026, opportunities generated by channel — evidence that contextual, multi-channel outbound still works. | | Skills = top 1% · Scheduled tasks = top 0.1% | AI adoption tiers | Mica's rule of thumb for where a GTM user sits on the AI adoption curve; most teams use neither. | | 37 | GTM/sales skills available | The collection of one-click-installable skills at amplemarket.com/skills that Mica urges operators to explore and adapt. | | 3:50 AM, weekdays | Product Hunt agent wake time | The scheduled agent that scrapes overnight launches, qualifies them for B2B fit, and routes them to the right rep before the team logs in. | | Thousands globally | Amplemarket customer base | The scale of sales and GTM teams using the platform, the source of Mica's pattern library of agentic flows. | | Millions of dollars / month | Amplemarket self-run pipeline | Amplemarket runs its own product for its entire sales team, generating millions in pipeline monthly across email, calling, and social. | | Most jobs since 2000 | US employment reference | Mica's counter to the 'AI job apocalypse' narrative — by his account, the world has never had as many people employed. | ## Entities mentioned - **Ample Market** (company) — Mica's company; the consolidated GTM platform (signals + contact/business data + multi-channel engagement) whose new MCP and agent layer power every demo in the episode. It runs its own outbound to generate millions in monthly pipeline. · https://leanscale-knowledge-hub.netlify.app/company/ample-market/ - **LeanScale** (company) — The host's firm; Anthony cites LeanScale's internal AI use — CRM+transcript diagnostics as a POC, transcript-fed project management, and account-health agents — as an example of raising quality and margin without firing anyone. · https://leanscale-knowledge-hub.netlify.app/company/leanscale/ - **Product Hunt** (company) — The signal source Mica's 3:50 AM scheduled agent monitors: it scrapes overnight launches, qualifies them for B2B fit, and routes the good ones to the right rep inside Amplemarket. · https://leanscale-knowledge-hub.netlify.app/company/product-hunt/ - **Anthropic** (company) — Maker of Claude, the LLM Mica drives throughout the demos via connectors, skills, an Amplemarket MCP, and scheduled tasks. · https://leanscale-knowledge-hub.netlify.app/company/anthropic/ - **OpenAI** (company) — Referenced as the maker of ChatGPT/GPT (and Codex); cited as the alternative frontier LLM users can wire into the same agentic GTM flows, with a new '5.6' model noted as just announced. · https://leanscale-knowledge-hub.netlify.app/company/openai/ - **Mica** (person, guest) — Founder & CEO of Amplemarket, the consolidated GTM platform with a signals, data, and multi-channel engagement stack now powered by an MCP and agent layer. · https://leanscale-knowledge-hub.netlify.app/guest/mica/ - **Anthony Enrico** (person, host) — Co-founder of LeanScale and host of The LeanScale Podcast. · https://leanscale-knowledge-hub.netlify.app/guest/anthony-enrico/ - **Claude** (tool, AI Assistant) — The AI assistant Mica uses for the demos, wired to connectors, skills, an Amplemarket MCP, and scheduled tasks to run agentic GTM workflows. He notes Claude will even interview you and write the scheduled-task instructions for you. - **ChatGPT** (tool, AI Assistant) — OpenAI's assistant, cited as the interchangeable alternative LLM ('you can do this on ChatGPT or Codex too') with the same connector-and-agent capabilities. - **Model Context Protocol (MCP)** (tool, AI Integration Protocol) — The Model Context Protocol — described as a way to connect your platforms into your LLM. Amplemarket's MCP lets Claude call its data and engagement layer from within a chat, so a prompt can enrich and act. - **Telegram** (tool, Messaging) — Where Mica triggers his 'Hermes' follow-up agent — a post-meeting message ('send Anthony a connection request') that hands off to an Amplemarket workflow agent; one message fired connection requests to a whole SDR team. - **Slack** (tool, Team Messaging) — Monitored by Claude agents for inbound-lead channels ('bot inbound') and high-priority deal alerts, which get routed into Amplemarket's inbound agent for suggested outreach and multi-threading. - **Outreach** (tool, Sales Engagement) — Named alongside Salesloft as the home of the templated, first-name-swap outbound Mica argues is now dead. - **Salesloft** (tool, Sales Engagement) — Named with Outreach as the sales-engagement tooling associated with non-contextual, mass-volume outbound being displaced by contextual, agentic outbound. ## FAQ **Q: What are the building blocks of an agentic GTM stack?** A: Mica describes four interchangeable 'Lego pieces': a very smart intern (the LLM, like Claude or ChatGPT), web search, a data platform, and a workflows agent that can take or suggest actions on your behalf. Any go-to-market play is a different combination of those blocks, so the real work is designing which recurring tasks to automate — not finding a single magic tool. **Q: What is the difference between a skill and a scheduled task in AI for sales?** A: A skill is a precise instruction set that sharpens an LLM's output and, for Amplemarket customers, installs in one click. A scheduled task is a routine that runs an agent on a cadence — daily, weekdays, or at a specific time like 3:50 AM — so the work happens automatically. Mica's rule of thumb: using skills puts you in the top 1% of AI users in GTM, and using scheduled tasks puts you in the top 0.1%. **Q: What is an MCP and why does it matter for outbound?** A: An MCP (Model Context Protocol) connects your platforms into your LLM so Claude or ChatGPT can call your data and engagement layer from inside a chat. It matters because it turns the model from a text generator into an operator: with Amplemarket's MCP, a single prompt can enrich a list, decide the next action, and take it — because the platform already owns the channels and sits on the CRM's context. **Q: Is outbound sales dead in 2026?** A: No — templated outbound is dead, not outbound itself. Mica argues the non-contextual, mass-volume, first-name-swap motion run in tools like Outreach and Salesloft no longer works, but contextual, signal-based outbound thrives. His proof: across Amplemarket customers since January 2026, opportunities split roughly 40% email, 40% calling, and 20% social, and the company drives millions in pipeline monthly running its own product. **Q: What is the two-agent pattern for AI-driven outbound?** A: It splits the work between two agents. A research-and-enrichment agent runs in the LLM, plugged into the web, to find and understand prospects. It then hands off to an engagement agent living inside the GTM platform, which is harnessed to your owned channels and CRM and has its own instructions for how to act. The first maximizes web context; the second maximizes CRM and channel context — together they produce action, not just insight. **Q: How does true per-prospect personalization differ from a mail merge?** A: A mail merge runs everyone through the same sequence and swaps in a first name. True per-prospect personalization generates the sequence from scratch for each person: the engagement agent knows whether you're already connected, whether you emailed or cold-called before, and what the CRM says, so it decides both the copy of each step and the channel each step starts on — LinkedIn, email, call, or text. **Q: Does AI reduce the number of sales and SDR jobs?** A: Mica and Anthony both say no. If AI lets a team go faster, competitive pressure forces you to reinvest that speed rather than bank it, so you don't fire people to move at the old pace. Mica's fastest-growing team is his SDR team, and he sees customers doubling and tripling BDR and AE orgs. The real effect is a widening productivity gap between operators who adopt AI and those who don't. **Q: How do you design an AI agent if you're not technical?** A: Use the 'smart intern' heuristic: ask what recurring work you would hand a very smart intern you just hired, and build agents around that answer. You don't have to write the instructions yourself — tell Claude what you want the agent to do and how often, and it will interview you (what to look for, who to target, how to deliver results) and write the scheduled task for you. ## Timeline - **00:00** — Cold open - **02:00** — What makes Amplemarket the GTM AI stack - **04:00** — The Lego pieces framework for agentic GTM - **08:00** — Skills + scheduled tasks: the AI adoption tier ladder - **09:30** — Demo — Pulling podcast attendees in one prompt - **12:30** — Demo — Mapping VP Sales at conference sponsors - **15:30** — Demo — Parsing a screenshot into a sequenced list - **20:30** — The two-agent pattern: research + engagement - **25:00** — Truly per-prospect personalization (not first-name swaps) - **28:00** — The Telegram + Hermes follow-up workflow - **31:00** — The top 0.1% — scheduled tasks that run while you sleep - **33:30** — The 3:50 AM Product Hunt agent - **40:00** — One Telegram message, bulk SDR connection requests - **42:30** — Is outbound dead? The 40/40/20 channel data - **47:00** — What the next 12 months actually look like for GTM - **52:00** — The ceiling has been ripped off ## Related episodes - **Ep. 95: Why AI Means More RevOps Hires, Not Fewer** (Jimmy O'Halloran, New Relic) — The direct thesis match — a productivity gain should trigger hiring, not layoffs; pairs with Mica's 'market forces you to reinvest speed' argument. · https://leanscale-knowledge-hub.netlify.app/podcast/jimmy-ohalloran-new-relic-revops-consumption-revenue/ - **Ep. 85: Why AI + GTM Engineers Can't Replace RevOps** (Tessa Whittaker) — The sibling argument that AI and a few engineers don't shrink the operating layer — the same headcount debate Mica lands on. · https://leanscale-knowledge-hub.netlify.app/podcast/tessa-whittaker-ai-gtm-engineers-revops/ - **Ep. 88: Why AI Won't Close Your Biggest Deals** (Michael Kiernan, Nextdoor) — A CRO's view on the limits of AI in enterprise selling — a useful counterweight to Mica's bullish agentic-outbound demos. · https://leanscale-knowledge-hub.netlify.app/podcast/michael-kiernan-nextdoor-ai-wont-close-deals/ - **Ep. 91: Why Outcome-Based Pricing Is a Trap for Most AI Companies** (Roee Hartuv) — Pricing-and-packaging counterpart for AI-era GTM, adjacent to Mica's take on how AI reshapes the sales motion. · https://leanscale-knowledge-hub.netlify.app/podcast/roee-hartuv-outcome-based-pricing-trap/ - **Ep. 89: Alex Wakefield (AcuityMD)** (Alex Wakefield, AcuityMD) — Another in-hub conversation on modern, data-driven GTM — pairs with Mica's signal-based outbound thesis. · https://leanscale-knowledge-hub.netlify.app/podcast/alex-wakefield-acuitymd-ceo-to-cro/ - **Ep. 93: Jonathan Hunt (HubSpot)** (Jonathan Hunt, HubSpot) — A GTM-platform-side perspective on AI in the stack — a counterpart to Amplemarket's agent-and-MCP layer. · https://leanscale-knowledge-hub.netlify.app/podcast/jonathan-hunt-hubspot-media-empire/ ## Full transcript _Machine-transcribed and not diarized; speaker attribution is inferred._ _Transcript only, as a separate file: https://leanscale-knowledge-hub.netlify.app/podcast/mica-ample-market-outbound-agents/transcript.md_ ### 00:00 — Cold open **[0:00]** (logo whooshing) Today I have a dear friend, Mika. He's the founder and CEO of Ample Market. I think one of the best sales and marketing platforms you could leverage if you're growing a business, you're trying to get in front of your ICP, I don't think there's anything better than Ample Market to use. We use it internally at LeanScale and we recommend it to a lot of our customers. And today, Mika prepared a special demo where he's gonna be walking through how to leverage Clod with Ample Market and what you can do with their new MCP. So Mika, super stoked. I can't wait to go through this. Ample Market already is such an amazing platform **[0:40]** for sequencing, data enrichment, finding your ICP, getting in front of people on multi-channel, LinkedIn, email. So I'm a huge power user of Ample Market already, but I can't wait to see what you put together and how we can do even more with AI. - Love it. Thank you for having me, Anthony. And again, my idea for today would just be to kind of like showcase a few of these AI or agentic flows in sales or GTM. Again, we're lucky to have like thousands of customers all over the globe and I spend a lot of time with some of the most just telling you these stories about like spending time with customers here in San Francisco and kind of like trying to understand **[1:20]** how a bunch of these companies are actually applying AI to their sales process, right? To their SDR motions, to their e-motions, to their customer success motions. And it's funny because like, and we were just talking about this as kind of like, if you look back like 12 months ago, like the way that like engineers were working was fundamentally different than the way that they're actually working today. And if you look back 12 months ago, the way that sales seems work was not so, so different than the way that they are working today. So like, I think there's like, we're making like the Delta is the changes like have not been that meaningful yet **[1:54]** when it comes to sales. Again, everybody's using a cloud or a LGBT, right? But they were already kind of like doing it. And I think skills were a thing that came out but you'd be surprised by the amount of people actually not using or not knowing what a skill is. And I think there's a lot still of like AI penetration, especially in the GTM org that needs to happen. And again, I bet you find this too, like, but there's a big variability in terms of like the AI savviness of a bunch of these organizations, even for like some AI first companies, there's still like a lot of variability there. So yeah, I'll shoot my screen and I'm excited to kind of like, ### 02:00 — What makes Amplemarket the GTM AI stack **[2:37]** my idea would be to kind of just inspire people and it's not at all import market related, like whatever like tools like you're using, like just want to inspire people to kind of like think a bit more about what can you accomplish if you have the building blocks that look like this. You have like a really smart intern, you have like web search. So like an intern that actually is capable of like searching the web for things. And then like, I'm assuming that you have like a data platform and I'm assuming that you have like a workflows agent, right? Some way of actually taking actions or suggesting actions on behalf of you. **[3:12]** And then like, we can go crazy, right? So to me, these are like Lego pieces, right? Different colors and we can just build anything which is sometimes tricky. Like when you go to the Lego store, you have like these sets of Lego pieces that are already like, you know, you want the Batman Lego piece, right? But I could send, I could tell you like, here's a hundred yellow Lego because the pieces and a hundred black yellow Lego pieces, but you wouldn't know what to do. And I think sometimes like, especially when AI and GTM, you get a bit into that motion. It's kind of like, okay, this is really powerful. Like, where do I start? Like, what should I do? **[3:48]** And I think inherently like you should, everybody should be spending time thinking about it. Like I think all companies should be just like, and I was doing this with companies like, but like just whiteboarding things like, where are the places? What are the things that are happening like on a recurring basis? What are the things that my team is actually doing that I can actually provide them like with a very smart agentic assistant and where they would actually bring my reps information, right? And again, just, I'm assuming from now on that like we know what cloud is or LGBT. And I'm assuming too that like, you know, like you have connectors enabled. ### 04:00 — The Lego pieces framework for agentic GTM **[4:27]** So meaning like your LGBT or your cloud are capable of like doing, they're capable of controlling your computer or going to your Slack and going into your like ample market or like into your data platform, into engagement platform and doing things on behalf of you. But let me share my screen. And again, like if none of this makes sense if you have questions, just, - No, no. I think that's perfect. And I think the audience is probably AI fluent, **[4:58]** at least definitely exposed. So I think going through that assumption is gonna be really good. And I think one thing just to tee up that I think would be good before going into some of these, just a quick overview of ample market. I know these use cases could be used with other tools, but I do think ample market unlocks quite a bit. And I think it'd be good to kind of ground how you're connecting with some of these use cases too. - Matt, and thank you. I'll give you the quick overview. So like ample market, like I think you described quite well before, but it's kind of like a consolidated platform when it comes to anything **[5:33]** that you typically need in your sales tax. So it's a signals platform. It's a data platform, like a business data. So like you have contact and business data, but it's also like a multi-channel engagement platform. So we're plugged into your email, we're plugged into your LinkedIn, we're plugged into your calls, and we can actually help you create sequences or steps that actually take actions on behalf of your suggest things to you. But the most interesting thing about ample market is kind of like if you have like really good data infrastructure and really good workflows or sequencing infrastructure, and if you put like a layer of AI on top of all this, **[6:04]** then you're truly like, it's so easy to start agentic flows, right? So ample market being like directly connected into your CRM, we have like maximum context about like what's happening in your CRM. We own all of your channels of communication, and we have a really strong and powerful database for business data. And so like then everything that you wanna build as like a GTM or a salesperson becomes so much easier. And so ample market powers agents, like, and I'll show you that in a second, but like you can actually set up like different agents that are doing things on behalf of you or that are just information agents, they're bringing things to you. **[6:38]** So again, think about ample market. If you want like a very simplistic way of thinking about it, it's kind of like a signal, a signals platform, a data platform, an engagement platform merged together, and then like capable of like, you can operate it the old school way, or you can operate it the new way, which is like, you can actually set up agents that are taking actions on behalf of you or bringing things your way. Does that help Anthony? - That's perfect, that's perfect. - Awesome. Before I go into these flows, I think one of the things that I really want people to, if you wanna become like a top 1% user of AI in GTM, **[7:16]** I think just like understanding what a skill is and leveraging skills can actually get you there. And again, I live in San Francisco, so like everybody in this town is AI-pilled, and oftentimes I still find people that are not like fully leveraging skills, especially like in GTM or in sales. So at ample market, we have like this space that is called amplemarket.com/skills. I would urge you to just go there and take a look at them. Again, these are the collection of 37 skills. And a skill just to quickly define this is just a subset of like more precise instructions that you're giving like your LLMs, so like be it Cloth or GPT, **[7:54]** and then asking like the LLM to be more precise in the way that they follow the instructions. So like you're maximizing the quality of your output. So like usually skills are kind of like tough to install, but at ample market, especially for ample market customers, it's just like you click the Cloth, you click install, and it's going to be like a one click install. So now like all of a sudden like this SDR comfortable decision skill is available to me. So just go into the skills page, take a look at them, and then you feel free to adapt these skills to whatever tools that you're using. But I think if you can take away one thing, ### 08:00 — Skills + scheduled tasks: the AI adoption tier ladder **[8:31]** just kind of like go into search for a GTM or sales skills and try to understand what they are and how you can actually leverage them. Okay, now let's go into a few specific workflows that I've seen like companies kind of like leveraging on a more recurring basis, especially when it comes to leveraging an MCP. And an MCP is just like a way to connect like your platforms into like your LLMs. So be like your GPT or your Cloth, but you're basically, you're giving like that LLM, the power or the context that exists in other apps. So Ample Market, we have like an MCP. And so basically you can call, you can ask Ample Market to do things via like your, **[9:12]** your MCP within, for example, a cloud conversation. And so here, for example, like I was just looking at the very simple use case here, which is this one, can you find the 10 last attendees of the Linscale podcast for each speaker, give me an insight from that podcast that is related to AI transformation in sales engineering, also provide an icebreaker, a fun fact, and short LinkedIn connection message mentioning your podcast, create a list on Ample Market. And then again, I did this. And then like, as you can see, like Ample Market was actually called in one of these flows where it's actually, we're enriching the data, ### 09:30 — Demo — Pulling podcast attendees in one prompt **[9:47]** we're getting the data from like the, we're getting actually the podcast transcripts and the attendees from Apple, like there are names, but there are no emails or no LinkedIn profiles. So, you know, you're getting all of the data and then you're actually enriching all of the data from here with Ample Market. And then the cool thing is like the outcome is that you go to Ample Market when it's done and you get a list like this. So you get a list of like Linscale podcast, like here's the first names, the last names and LinkedIn profiles, all of this data fully enriched and validated already for you. Again, as a person, this took me no time. **[10:19]** And now I have like two cool things here. It's just kind of like an intro, your Linscale episode on pipeline versus revenue was sharp. The data point that too much low quality pipeline drag when it's down is one of my teams, more teams need to hear, like that's another one. Like you have like a little insight here. But again, this is like the type of things that are very easy to do once you have like a smart assistant web search, right? Because here I can actually leverage a web search to go into like the Linscale like list of podcasts and then Ample Market MCP because like all of these contexts like Ample Market needed to know that like, okay, **[10:53]** this person that works at this company is this person that is like this one email. So we actually found the email, we found the LinkedIn profile and now we're providing you all of this information here. And again, as a rep now you can decide what you wanna do with this. It's kind of like just a very simple list building use case in here. - Yeah, and this is really spot on. I mean, if you were to try to do this on your own, you would probably go to the podcast if you even knew it was there 'cause I'm sure you could prompt to say, "Hey, what would be some interesting ways to connect with these people?" And then you can find that. **[11:28]** Then you'd have to go figure out who they are, then go try to feed them into an original model, but to be able to do it all in one prompt, that's pretty impressive. - And again, this is very basic, right? So because like, I'll show you how this becomes a lot more interesting, which is kind of like, now you have the podcasts, you're able to find the people, you're able to infer who the people are, and you're able to get their contact information. But then like, what you wanna do with this is potentially you wanna take the next step, which kind of like take an action or suggest one action. You might want to map these across like your own CRM, right? **[12:04]** And so like, some of these accounts might be owned by me, some of them might be owned by my team, and I might want to not send them to me, but to the right account owner. And by the way, I might want to take a look at lean scale attendees on a weekly basis. So instead of having to run this every time as a one-off thing, I want to run it like recuringly on a Friday afternoon, for example. And I'll show you that in a second. The other cool use case, and I think, again, that we leverage and we really like to do is just kind of like we sponsor, we are at events very often. And one thing that we like to do as a sales team, ### 12:30 — Demo — Mapping VP Sales at conference sponsors **[12:44]** we like to, again, map the floor, map the sponsors, understand who the sponsors are. And typically for the conferences that we go to, the sales leadership team will be there. So there'll be someone from VP of revenue, CRO, VP of sales. So one thing that we actually love doing is this thing, like, okay, I'm going to this conference that's happening in Europe that is called UNX 2016 Europe. And I like to get the contact information of the VPs of sales, sales directors and CROs for their top 20 sponsors. Find this and enrich them and add them to all these on Apple market. So like here, again, I know that there's like a page **[13:21]** for the conference, and I'm actually sharing that page with like my assistant, I'm telling my assistant go there, fetch all of the sponsors, which typically will be a logo. So you get a logo for the sponsor. So you need to convert that logo into like a name of a company, and then like you need to infer that that company is actually a specific business URL. And then from there, you actually need to go and find like VPs of sales, VPs of revenue, GTM leaders, and then add them to a list. Again, you can do this manually or like within a prompt here, you can actually then go back to Apple market and you can see that again, we have like their top sponsors **[13:56]** here and look like director of sales, VP of sales, VP of enterprise, GTM sales director. And all of these are kind of like the companies that are the main sponsors. And these are the potential sales leaders that they're going to likely have at that conference. Very simple one, but again, one that would actually take a while to execute. And again, like this is just to give you like some kind of like the building blocks of like, okay, web search allows me to go anywhere on the web and get context of what's happening there. The agent allows me to then parse that information and feed that information into like this ample market MCP. **[14:37]** And then that actually allows me to convert that information into actionable insights or into actionable data that I can then later on put into a sequence or send to my reps or send to my CRM for all intents and purposes, if that makes sense. - Yeah, and to be able to do all that in one flow is really, really difficult. 'Cause normally you'd go like find the list, then you would try to get the list into another platform and then enrich it, and then take the enrich list and go put it into a sequencing platform. I mean, there's enough friction in there to where you probably won't even run some of these plays **[15:16]** because it's like, ah, it's gonna take too long for me to even get to that point. But to be able to do the whole thing from Claude with the backend of ample market, giving you all the data that you need, that just really unlocks the ceiling of what your outbound efforts could be. - I completely agree with you. And again, I'm taking this example here, which is kind of like, look, you have a list of like, just like this is a janky one, but like I went here, took a screenshot of like a few Twitter followers for X followers for ample market. And then I can actually tell like Claude, can you please find the people in this list ### 15:30 — Demo — Parsing a screenshot into a sequenced list **[15:58]** that follow ample market, they use ample market **[16:05]** to find their LinkedIn and their email address. **[16:13]** Create a list for ample market X followers. **[16:21]** But again, this is just to showcase another, a bit trickier principle is kind of like, even if you have like data like an image, like this like really smart agent can actually parse that information. Then there's going to be like non-perfect data here, right? So like you have a non-perfect data you have, and it's really tricky to sometimes like parse like information from like a place like Twitter or like people might use different data points and what they have in their LinkedIn or in their email address. But like with ample market, you're kind of like going to parse this with Claude and then push that data into the ample market MCB. **[16:54]** And there's like a lot of complexity here. Like you have first names and last names and maybe there's like a company name. So there's a lot of like potential like combinations of searches that you can do here. But this is another cool example of something that you can do today. And I'm just showing this one because it's an obvious one, a tricky one that most people actually wouldn't do. But like now you just let the ample market parse. Like, so it's actually doing the search here. And then basically it's pushing this all into like the ample market MCP, which has access to like a search that can ingest like names and company names, et cetera. **[17:28]** And then eventually like I'll have a list that is created for me on like then, on ample market, sorry, with their LinkedIn profiles, et cetera. So let me share with you. But yeah, we'll just wait for it to pop in here. But I just wanted to show that with that one too, which is basically very interesting because like when we go to conferences, we have this other thing that allows us, we take pictures of badges and then like, I say like I'll connect with this person and then ample market is, that's actually an automation. Like we infer like person name, last name, company name. We find the person and it connects automatically **[17:58]** after the person swam by the booth, for example, which is a pretty interesting use case. But as you can see here, like we're finding ample market is finding these results. It's actually finding everything. It's enriching all of these people. And then it's going to be adding them to the list. But yeah, that's another cool one, Anthony, that I wanted to showcase here because it's not an obvious one, going from like a stale picture into like something that is like context and contact rich. - I think you've really pulled together some of the more difficult parts, especially I love the events one where you're taking pictures **[18:33]** that way you don't have to have a separate app or you're taking time scanning a QR code. It's a quick picture, way more comfortable experience when you're interacting with people in person. But it's stitching together a lot of these really difficult things to get the right list and dataset to. Now this is where I think a lot of this tends to break down. How do you, you've enriched the data, you have the right list, you have the people you want to get in front of, you have the context. Now how do we put that to action? What do we do next? - Yeah, very good question. So like these ones are kind of like the list building ones because it's like, **[19:11]** I see a lot of people using platforms just to list building. I'm telling, what I'm wanting to showcase here is like, once you have a really good data MCP connected to like your judge of duty or your claw, like, you know, list building costs are very easy and it's kind of like very basic, but this building can be very sophisticated. But to your point, I agree. Like how do you go from here now to an action? And that's the other magical part of AMPL market. I'm actually going to just open up AMPL market, but I was telling that, I was sharing that AMPL market is as the capability of actually ingesting a signal and then like suggesting actions **[19:43]** or taking actions on behalf of users. So the way that you connect now Claude to AMPL market, you can actually, I could say that we can go there and say like, oh, now build a sequence and like write this email, et cetera, and then send to AMPL market, right? That's one thing. Or we could just connect the Claude agent that is running the parsing and creating the list and say like now send to this other agent, there is an agent that lives within AMPL market. And then these two agents are actually communicating. So like this one did the research enrichment and then I was actually sending to like another agent on AMPL market. **[20:17]** And that agent within AMPL market has specific instructions for how to act on the data that the other Claude agent sent to it. So like you have like, it's almost like you have two agents here communicating and every single one of these little things here, as you can see, is a different agent. So for example, like I was, again, we talked about the lean scale one. So Chris here went into the lean scale podcast. And so like Chris was fed into this other AMPL market agents where you can see here that, you know, this agent is like, I don't want to sequence. I just want generic task for it. Tell me why this podcast and this person would be interesting ### 20:30 — The two-agent pattern: research + engagement **[20:54]** for AMPL markets. And so like right now, Chris went into the lean scale podcast. I actually received this on a daily basis for new alerts and information. It's actually giving me information for the accounts that are under my name. That's an example. But I have like other, it's like, I didn't show you this one, but like I have one that is actually monitoring product hunt. It's like the way that you can actually monitor like a events page. You can actually have one that is monitoring product hunt. And so like here, for example, I have like, I have an agent that is actually monitoring product hunt launches. And then here, Artem, like it was, **[21:30]** his company was launching something. And again, same thing. I'm here kind of like getting information about like what the launch was, **[21:38]** how many upvotes there were, what were the most interesting things in the comments. I have another one, for example, that is connected where like Claude is looking at a Slack channel for inbound leads that my SDRs are working. And then it's actually giving me, it's actually giving this agent here, which is called my inbound agent. So like, again, you have like a Claude agent that is running on Slack. And then it's actually going and sending that information to the AMPL market, my personal AMPL market inbound agent. And it's actually saying, "Nick, Nick here from our AMPL market." So you're coming to check us out, like happy to help. **[22:12]** Let me know if you want to chat one on one. So that's like Nick here at Trotto was, you know, scene coming inbound is actually engaging with someone in my SDR team. And now I have like another agent that is actually suggesting that I take this specific action. And now like, I have this other agent, many, many different, but this was actually one. It is also monitoring Slack for the most important deals, like that are coming in. And it's actually suggesting that actually multi-thread with C-level leadership within those accounts. So for example, this was actually like, again, senior director coming in, CRO getting a connection request from me. **[22:51]** So there's a lot here, but connecting Claude to AMPL market in this case, because AMPL market is capable of like data and engagement and creating its own agents. These things become very simple and very easy, right? So all you need to do is to kind of like create a new agent and within that new agent, you just like give it instructions. And so now what you have is kind of like two agents working in tandem for you and either taking actions on behalf of you or suggesting actions that you should undertake. And for me, like every day I come here, okay. And I, all I do is kind of like I say yes or no to these tasks, if that makes sense. **[23:35]** I have more, but like I wanted to pause here for a second to see if this overall makes sense. - It makes a ton of sense. And I like that you really put the keys in the hands of the person who is doing the outreach and doing the outbound because you can say, "Hey, yeah, run it fully autonomously, like go if you want to. There's checkpoints you can have where you keep the human in the loop to make sure, hey, I wanna triple check. These are super important outreaches. I wanna make sure that these are going to the right people or tweak the messaging a little bit in sequence before shooting send. But I also wanna highlight how personalized this is. **[24:14]** This isn't saying, "Hey, we're gonna run a sequence that has seven steps and we're gonna run everyone through it and we're gonna personalize it by changing their name or adding a little tidbit." No, no, no. Each person is getting their sequence started from scratch. Should they be hit up on LinkedIn first? Should we send an email first? Should we do a task for this first? Should we build an asset before we launch a sequence? I think some people listening might not have an appreciation for exactly how personalized this is and how signal relevant the sequence becomes based on what's actually happening. - Very good point. **[24:52]** That's like, thank you for bringing that up. I think it's like, but this is an agent. So the agent can do like the instruction that you give it. You can act, again, think about this agent also being another smart agent that knows whether or not we are connected, that knows whether or not like I emailed you in the past, that knows whether or not I cold called you in the past and you picked up the phone. It's like whenever like the information comes from like the web and the cloud agent into Ample Market. Ample Market, since like we own like these channels of communication, the agent can then suggest something that is truly unique, right? ### 25:00 — Truly per-prospect personalization (not first-name swaps) **[25:22]** So like it can say like, okay, like if I sent an email to this person, I wanna start with an email. But if I cold call them in the past, I kind of like wanna start with a cold call because or even a text message, right? So if I have their phone number, if I've engaged with them in the past, so like now you have like, not only like the content of every single step is truly bespoke and unique, but like even this, the where you're taking actions is unique and so this is like another one, like this is a cool one like, but a quick example that I wanted to showcase here is like actually like a referral from a friend like, hey, George, can you see it on me? **[25:56]** So just as I reaching out to you and across the growth team, like you mentioned the growth engine, you're building a curse. It's like, this is like, this is actually, this context here, plus you overlap with (indistinct) there's like a lot of context here that I just really quickly gave it to like my agents. Oh, some just remind me to connect with George because like, Kirti told me that to reach out and this was the topic. And so like, I told that to, I told that, actually I told that to a telegram to my Hermes agent that then I sent that to my ample market agent that then created this very simple introduction sequence here **[26:30]** from me with like an email and a connection request. If yeah, if that makes sense. And I wanna show a few more advanced things here, but like, yeah. - Does make sense. And just the other thing to highlight is, you can leverage so much more context where a typical like, oh, going into do a sequencing platform or something, it doesn't know you, you can't add the company context. You can't add a whole voice profile of how you word things and exactly how you write things. So it sounds like it's you. It doesn't sound like it's written by AI. There's just so much more you can do. Ample market has always been really powerful, **[27:11]** but now opening up these AI agents on top of it is just, it's really exciting. - And to that point, like I think this is my production. This is my real account. So I mean, I'm worried about like showcasing things from customers, but I think I can show this one because like the opposite zero is like, there's no much here. But here, like here's the purple things here. These are like context. This is a lot of context that is coming from CRM and from all interactions that have happened like with this account and this specific person in the past. So like, so this is a very dense and very rich context here that you can then push on to whatever you're doing **[27:52]** in terms of like an approach. And I'm not going to open the other ones because it's customers here, but like, but all of these like account insights are truly unique and bespoke. And so like the agent that is building the sequence, I can tell it because I'm mapped with the CRM. Take a look at the CRM, AE mantra, SDR mantra, or pass, close, lost, reason, and feed that into your creation of the story, the why for this person, right? So like there's a who that comes with a specific reason. We propagate that who with the why. Why is Chris here, right? And then like, do we have more context? We were talking about data before, ### 28:00 — The Telegram + Hermes follow-up workflow **[28:29]** but like, do we have more context that I can give it? Yes, like, oh, they were part of a close loss opportunity, for not the right time three months ago. So you actually like, and you talk with Mary, and so mention Mary to Chris. So like, that's something that you can very easily prompt the agent to do here. But like, one is kind of like you have an agent that is plugged into the web and doing things over there, which is like nice. And then like you have this other agent that is harnessed into like your communication channels, like all of these channels that you use to communicate, and then like your CRM. So it has a lot of information. **[29:06]** And so you can actually harness this other agent to be maximally productive and contextual when it's building like a sequence and the content within each step. And then again, like not sure which AMPL mark, but that's why AMPL mark is so powerful, because again, you can truly combine all of the data together. **[29:27]** - I love it. I love it. How else can we take advantage of cloud or open AI's connection to AMPL market? - Yeah, again, all of the things I'm showing you here are, you can do them like on chat with your open AI codecs, et cetera, but like, I'm just like a cloud user. So that's why I'm kind of like showcasing these things from cloud. But the one thing that again, like going more into like a broader, more general thing, things that you can do today that you might not be doing and likely you're not doing. Like again, going back to the fact that I was like, again, if you're using skills, you may be top 1% AI users. If you're using scheduled tasks, **[30:10]** then they are top 0.1% AI users. And so like, I'll just describe what scheduled tasks are. And then I'll showcase like in a second, what is inside them. But a scheduled task is just like a routine or like it's just a little set of instructions that you want your cloud or your chat GPT agent to run ever so frequently. So for example, this one runs every day here, like at 6 p.m. This one runs on weekdays only at 9 p.m. Some run like once per week, right? So this one runs every Wednesday at 3.58 a.m. Very specific time. But what is in there in here is kind of like, you can have your computer like shut down, like not completely shut down. **[30:57]** Like you can just like switch off your display and then like your cloud agent will wake up at 3.58 a.m. And it's going to execute those instructions for you. And then it's going to send you information, right? Or do things on behalf of you. So I have a bunch of little agents here. So like the one that was just showcasing you before like the product hunt one. So like on weekdays at 3.50 a.m. Like I have an agent that wakes up, goes into product hunt, takes a look at the launches from the past 24 hours, gets all of them and reaches all of them with ample market MCP and then throws them into the ample market agent. So like now you have like these agents ### 31:00 — The top 0.1% — scheduled tasks that run while you sleep **[31:37]** like by the time I wake up, I now have like a list of all of these product hunt launches that make sense to me because like the agent is also like applying some sense of like quality and discarding a bunch of them. I have this other one like revenue podcasts. Like so like I'm not only for lean scale but 30 minutes to presence club GTM now. I have an agent that like every Wednesday is actually monitoring the latest episodes from these podcasts. And it's bringing me information, right? And then sending that information to my agent that can then like within ample market either route them to me or to the right rep within ample market. **[32:16]** But I want to show you like what that looks like. So for example, this is just like the set of instructions that you were doing with that agent. And that agent becomes now, it's kind of like an autonomous agent, right? So that it's running every single day at a 3.50 AM. And it's again triggering a set of instructions that I gave it here. And so the steps is like scrape yesterday's product hunt launches, qualify each launch for B2B fit. So I don't care about all of them. I just care about like B2B extract for unique launch insights, right? So, and then you put them into dynamic fields. You send those into a list within ample market. **[32:56]** And then you feed them into like the ample market. Again, this is actually sanitized but in this you shouldn't worry about this. But you feed that into the ample market agent which is kind of like this. It's just a specific end point here. But this, it seems complicated but it's actually very easy to create because Claude will, if you say, "Hey Claude, I want to create an agent that runs every day at 3.50 AM that would be connected to my ample market. And it's actually monitoring like product launches. How can I do that?" And so all of this set of instructions, Claude will actually create for you by just asking you questions. He'll ask you like, ### 33:30 — The 3:50 AM Product Hunt agent **[33:32]** "What do you care about in those launches? What types of people would you like to go after? How should I send the information? Do you want this to be written to a list? Do you want this to be acted on a sequence?" So you can, you know, this feels complicated but it is not. It's actually very simple and Claude can help you do this very easily. There's this one, inbound agent scan. This is actually like, again, I really want you to think about the building blocks. There's a Claude or a chat GPT agent that is connected to the web and that can actually have access through connectors to the apps that you mostly use in the business. **[34:10]** So we have this like channel called bot inbound. And so like within bot inbound, I want this agent to monitor all of the opportunities or all of the demo requests that happened there. And I wanted to kind of like then pull that in and information and reach those companies and then send them to me for a specific action that I will undertake. And then we saw what that action looked like in the ample market dashboard, right? So like, and for some of these agents, I would actually feed them into like an automated workflow. I don't, sometimes like connection requests, I don't care. Like just, you know, it's just a connection request. So run it for me. **[34:49]** I don't need to be waiting to pre-approve it. But when we're actually doing like some writing, I love to kind of like be able to preview the output of like both of these agents work. But again, I would say like, if you think about it, go into like a schedule task and all you need to do is kind of like, you can create with Cloud Sys, as I was describing, you can create it with Cloud and Cloud will ask you when you want to run it, how frequently you want to run it. And it will help you write all of these instructions. You don't even need to be thinking about like these specific instructions, but I would challenge you. **[35:25]** Like, so if you're monitoring like a webpage as a ledger for new entries, and if a new entry is usually correlated with you having to reach out to someone at that company, you should just build a Cloud schedule task. And you say like, "Cloud, monitor a webpage so-and-so." Find like the name of the company there and then feed that into like whatever data MCP you're using and, you know, into and enrich that and then send that to my engagement MCP. I don't know like what people are using there, but this is super easy to do today where you can just go in here and then schedule that task to run for you every day at whatever time. **[36:02]** This is actually the important point for me to just like grasp here is like, I see some companies kind of like wanting to feed their own agents within their platforms, right? So like, and kind of like having the reps move out of Cloud or LGBT. And I don't think that is going to happen. I think you always, I think reps will always want to maximize the intelligence of their agents. So I think like if you think about like Cloud or LGBT, they will always be launching the best models. Like today, LGBT announced like 5.6, right? Different versions which just seem to be like, kind of like on the benchmarks, pretty powerful. And I think if you want, **[36:43]** I think users will always want to be running those maximally intelligent agents on behalf of them because they always have like access to like connectors and different types of workflows like loops and goals, which again, these are more advanced things. But and then like plug them into like their existing flows. I think I see so many people nowadays working from within Cloud and LGBT, but it's like, how can you get Cloud and LGBT to work for you without you being there? And hopefully that's like, it's something that you can actually get out of this specific example, which is get these agents to work for you. **[37:24]** The important thing is to kind of like just sit down with a white paper and then just try to understand and think about like the things that you do on a daily basis that you wish if you had someone working for you, if you had like a very smart intern that you just hired, what would you ask them to do? And I would actually operate that way with like your agentic flows. I think these can be super fun and they're not that hard to set up. Again, you need a platform for data, you need a platform for workflows. Again, I wonder if he works very well for me. So, but yeah, hopefully that makes sense. And again, I have many other examples in here, **[38:02]** but I think I don't want to overwhelm people so much, but again, like I'll just show you a last one, which is I usually when I have a meeting with someone, I would, you know, after that, the meeting like in person, for example, I have like a little telegram, Hermes running, let me see if I have my telegram. I have like, you know, my Hermes agent here running on telegram. And so then I would actually say like, you know, I just met with this person, for example, I just met with Anthony. Can you actually send Anthony like a connection request on my behalf, right? So we can actually then like, I usually call that like a follow-up post meeting. **[38:48]** And so I was actually like doing that for you, Anthony earlier, and it was like, okay, cool. Like I want to send Anthony into this workflow. Or this one is actually like, again, I just met with the new SDR team at Omni, or I don't know, this is the resolve, sorry, it's resolve.ai and I want to send all of them a connection request. And I was actually on the go and this is again, like this is what it looks like, but then like this sent like the information to a Hermes agent that then they communicated with this workflow agent and it started like, you know, running all of these things completely automatically for me. **[39:22]** Again, this is what I feel like the future and the gigantic flows are for us, right? So it's kind of like, how do you get these things to work by one, you pre-programming them to do the tasks that you care about, or you kind of like, when something happens somewhere, trigger the agent, right? So, you know, you could actually say like, every time I send a Slack message to this channel, monitor that channel, like, and then do something with it. But yeah, these are a bit more advanced workflows, but hopefully they kind of like inspire people to think about how to set up like agents in their daily workflows and how you can get these agents to truly add value ### 40:00 — One Telegram message, bulk SDR connection requests **[40:02]** to your sales motions. - Yeah, I think you did an excellent job mapping out art of the possible. And I think one of the pieces of advice that you baked in there, if you don't know how to use AI, just ask AI, and it's gonna give you step-by-step instructions of what you need to do, and it'll help you build whatever you wanna build. I think one last question before we wrap up, I think there's a lot of noise saying outbound is dead, there's too much noise, it's gotten too messy. What's your take when you're hearing that? Because it doesn't sound like that's the case for you or your customers. - No, at all. I think the part of outbound that is dead **[40:47]** is like the templated outbound, like what people do in outreach and sales loft mostly, right? I think using a thing like, hey, first name, I noticed like that you are at company name, let's do a call, like something like this, that's dead, right, like it's not how you should be doing outbound. Nowadays, you can be a lot more sophisticated, but sophisticated, I mean, like you need to be a lot more contextual, right? So for all of these things that I was showing you here, there's always a reason why these things are getting triggered. And they're usually like is more context that it's fed into the ample market agent **[41:20]** to make it a maximally contextual conversation, right? I think like the part of outbound that is dead is kind of like this non-contextual, like mass volume outbound. Although I would also challenge the fact that like a lot of teams spend a lot of time just paralleling on the phones. And that's again, something that we'll see how that depends out in the near future. But outbound is not dead at all. I'll give you a specific data point. For ample market customers since January, 2026, I was just running the numbers yesterday. If you take a look at opportunities generated by channel, it was 40, 40, 20. So like it was 40% email, 40% calling and 20% social **[42:02]** out of all the opportunities. And again, like these shows are like, you know, sometimes people think that the email channel doesn't work. It does work if it's done well, right? So like it's works when it's like, again, context reach and you're reaching out the right person, the right time through the right channel. Again, this is just, it's almost like, yeah, duh. But it does work when it's done well. And I think like, again, if you think 40, 40, 20, just running the numbers, they're gonna like... But the 20% of opportunities from social, social works really well. It's just because like you're kept in social and you can only send so many connections. ### 42:30 — Is outbound dead? The 40/40/20 channel data **[42:37]** So there's like you're more kept by volume. But outbound is not that at all. Like again, we generate millions of dollars in pipeline. Again, we run ample market. We run ample market to do sales for ample markets. My entire sales team, what they use is just ample market. And again, we generate millions of dollars in pipeline every single month from all of these channels, right? So we do leverage all of the channels, email calling and social selling. So again, I'm happy to prove people wrong. I have many examples of like very happy customers here. They're like, you know, use ample market for their upbound motions. But yeah, it does work. **[43:18]** - Yeah, I think lazy outbound is dead, but really good signal based because that person actually has a problem that your solution actually can help with or you have an offer that's interesting for that person. People connecting to find the right solutions for each other probably is never gonna go away. So Mika, I can't thank you enough. You literally never disappoint. These were amazing examples and I can't wait for our audience to see them. I can't wait for our customers to see them because we have a lot of mutual customers on ample market and then our team is using ample market as well. So just really excited to get the art of the possible out there. **[44:00]** And if anybody is watching this and is looking for a signal data and engagement platform, I highly encourage checking out ample market and then leveraging all the new AI capabilities that Mika walked through. - Thank you, thank you so much for having me. It's always so much fun talking with you in this crazy because I think we talk every, I think we did one of these like maybe like 18 months ago or 12 months ago and it's like the world is shifting and changing so fast. It's such a fun time. I wonder where we will be. Where do you think we'll be like in 12 months just like wrap this up? Like, you know, like it was hard for me to imagine. **[44:37]** I failed on the, I mean, it was really hard for me to imagine that engineering would be so changed. Like I was, I thought like, yes, they would change a lot but not as much at it. Like the job changed kind of fundamentally. What do you think? And it's like not specific to sales already. But like, where do you think this will all be with fables and mythos and 5.6 models kind of like scaring people like where do you think we'll be in 12 months? - It's a great question. I'm having a tough time forecasting, you know, 30 to 90 days out. **[45:13]** I think what's interesting, if you look at what the current capabilities are today versus the adoption, like you mentioned, hey, if somebody is running a routine you're probably in the top 1% of AI users. So even when the models get significantly more powerful I really think there's gonna be a disparity of productivity that the very curious and A plus players the people who are seeing this and taking advantage of it and running with it they're already leaving people in the dust. And I think that's gonna expand. It's almost gonna be this like inequity of productivity and capability that certain people will have **[45:54]** because they've invested the right amount of time into it. Now that's one aspect. I think the other is you still need to have taste of what good is. And I think you can't fully AI your way to doing that. You still, like you mentioned, hey, if you had the smartest interns in the world you still need to tell them what to do. And you still have to have a good idea and a good vision. And I think the people that have vision and strategy and understand what's relevant will just get a lot of productivity gains out of it. But I don't know, I don't see the big headlines are like, oh, is there a job apocalypse? I'm not seeing that. I think the main thing is, **[46:42]** wow, if I have a 10 person engineering team and they're a hundred times more productive, like great. Now the economics of hiring more engineers might even make more sense. So if you can just, if you can keep the productivity of a team going why would you slow down? - And the market will force yourself. Like the thing, I think the market forces you. Do you know what I mean? Like I think sometimes people, if you can go faster you're not going to fire people to go at the same speed that you were going before. Because all of a sudden like within the market your competitors will go faster. So like you don't have the luxury to your point ### 47:00 — What the next 12 months actually look like for GTM **[47:17]** like this inequality like that you're describing. If you're not doing, if you're not going fast like your competitors are going to outpace you. And so that's the, I think that's how the market stays in equilibrium. It's kind of like, no, we can't let people go. We just need to make people a lot more productive with like all of these agents. Because like naturally like all of our competitors are going to want to go faster. So that's like, you know, and by the way like the narrative is broken because we've never had, I think the only other year that we had like more people employed at least in the US was 2000. **[47:48]** So like we've never had as many people like with jobs in the world. But your point is interesting. The point of like the inequality, which is like interesting like yeah, yeah, I don't, I'm not worried at all for like sales jobs. I'm not worried at all for sales people. I can, there was like 24, 25, there was this narrative of like ASDR, like the sales people is gone. It's like, I don't believe in that. We never believed in it. We actually always took like the stance of like human as well as AI. But I am, I'm not worried. You know, my fastest growing team has been my SDR team. Like at least like in the GTM side. **[48:20]** Like, so like, I don't, I don't see what and I see that in customers too, right. I see customers doubling, tripling their, their, their BDR and their E orgs. And so like, yeah. - Yeah, I mean, if I'm looking at lean scale, we've absolutely implemented AI and everything we're doing. I'll give you some real practical examples. Like we leveraged before we even engage with a customer, we're leveraging the transcripts as well as a connection to their CRM to do a full diagnostic and say, we know what good GTM ops looks like. So we can fully compare their whole setup to what an ideal setup would look like. So we're definitely getting deals done quicker. **[49:04]** That's like our POC, if you will. And then we also leveraging the transcripts to do all the PM work. So now our architects really just have to have meetings with the customers. Those get automatically pushed as like projects and tasks. And then we have agents monitoring the health of our accounts and everything. We didn't fire anyone. We just increased the quality that we're delivering and we're passing that on to the customer. But our gross margin maybe increased a little bit, net margin too, a little bit, but not like, not to where I could say, oh, we can run the same amount of business with half the team because the expectations just increased. **[49:46]** They're like, oh, well, this other team has access to AI and they're able to do this XYZ. So why can't you do that? So it's like, we just had to step up in our quality, but the market just normalized everything. Now just people expect more. Like nobody's gonna expect you that you're gonna work less either. It's like, no, now that I can do unlimited work, there's more work to be done. It just completely ripped the ceiling off of what was on top of work. - And people are insatiable. We always want more. I guess never like, oh, I'm fine. Like this is good. No, no, no, if I can ask more from like vendors, like I always want more, go, I need this, I need that. **[50:25]** Like people always want, you know, like people are insatiable. They always want more, right? So you, so like, just like, and we're all people selling to people. So like in the end, like, hey, we're insatiable. I will go, let's just go faster and we'll find ways to stay busy. I think it's such an interesting time to be alive right now. Again, then we were talking about AI and GTM, but I think like AI in everywhere, right? So AI in your personal life, building your personal lives, all of the same for your family. It's that we talk, I think at some point like in the past, we talked about this, I think it's like such a fun, a fun time to be alive. **[50:58]** - I think it's a really fun time. I have a really positive outlook on it. I think it's gonna create a lot of opportunities. And I think it'll put humans into their roles where they can do their best work. So we'll see, time will tell. I know we'll do another one of these. And when we hop back on, let's see if we change our mind a quarter from now. But in the meantime, I'm looking forward to what's up. - Thank you so much for having me, Anthony. It was always fun. - Thank you, Mika. Appreciate it. --- _LeanScale Podcast Knowledge Hub. Free to quote and cite with attribution to The LeanScale Podcast (https://www.leanscale.team)._