Discover
GTM Vault
GTM Vault
Author: Rick Koleta
Subscribed: 0Played: 0Subscribe
Share
© GTM Vault LLC
Description
Revenue architecture for AI-native GTM, from the operators who ran it. The GTM Vault Podcast, hosted by Rick Koleta: founders and revenue leaders on why pipeline stops compounding and what they rebuilt. Every episode is written up with its full transcript in the Library at pro.gtmvault.co/library, free, no sign-in. Show Me Your Stack and Coffee Walk are on YouTube and gtmvault.co, and every tool named on GTM Vault is on The Stack Index at pro.gtmvault.co/stack.
www.gtmvault.co
www.gtmvault.co
54 Episodes
Reverse
GTM Vault is a breakdown of how high-growth companies design, test, and scale revenue architecture.A conference booth costs $90K to $100K and returns a badge-scan list nobody wants to work. A $500 Amazon voucher buys a 30-minute demo from a buyer who cannot name the company 2 weeks later. Sachin Gupta built HackerEarth, then started Breakout in 2024 with $3.25M against competitors that raised many multiples of that. He joins GTM Vault for episode 55 to explain why his team buried cold outbound in a hearse at Inbound for under $6K, how each stunt is measured before it runs, and why an inbound agent company spends its brand budget on things software cannot do.About BreakoutBreakout builds an AI inbound SDR that engages website visitors in real time, answers product questions, qualifies intent and books meetings, so mid-market B2B companies convert more of the traffic they already have. It was founded in 2024 in San Francisco by Sachin Gupta, previously founder and CEO of the technical hiring platform HackerEarth, and Hitesh Aggarwal, who spent more than a decade building AI products at Google and Citi. In October 2025 it launched Blocks, which turns static website elements into personalized, AI-driven interfaces, with CaptivateIQ, Barti and HackerEarth among its named customers. Breakout raised a $3.25M seed round led by Village Global, with Recall Capital and Z21 Ventures participating.Watch or listen now across YouTube, Apple Podcasts, and SpotifyEpisode highlights(0:00) Attention Is the Scarcest Commodity(2:40) What HackerEarth Taught Him About Focus(3:40) Building Into a Loud AI SDR Market(6:14) Why Breakout Killed the Complex Demo(7:48) A Funeral for Cold Email at Inbound(12:26) The Lego Stunt and the Attention Problem(15:39) Brand Activation, Not a Pipeline Target(18:51) How to Put ROI on a Stunt(34:33) Outbound at Scale vs the Inbound Shortlist(46:46) The Website Becomes the CloserWhat you’ll learn:* Why a $500 Amazon voucher for a demo buys a meeting and none of the attention, and what Breakout sends instead* How a staged funeral for cold outbound went from idea to team commitment in 3 hours and cost under $6K, against a $90K to $100K booth* The pre-launch math: a $500 to $1,000 cost per qualified lead means 6 qualified leads covers a $6K stunt* How $5 to $10 coffee vouchers turned a fuzzy brand stunt into a countable metric, people posting on the day* Why the funeral produced 20 to 25 leads at the event and another 8 to 10 inbound queries from social, with no pipeline target set* Why limited-edition Lego plus a LinkedIn and email follow-up hit a 30% activation rate where generic gift cards burn budget* The cardboard-sign stunt that flopped in San Francisco, and why shared office towers killed the foot traffic it needed* Why Breakout would cut paid ads before it cut stunts, and the one condition under which it would not* How 60% of Breakout’s closed-won comes from inbound, and why its own site has no form, only the agent* Why a 5-year-old press release on a customer’s site can lose a buyer at peak intent, and what onboarding now excludesKey takeaways1. Attention is the constraint, and a gift that is one of many buys none of it.Gupta calls attention the scarcest commodity in B2B, and the voucher story is his proof: $500 for a 30-minute demo, and 2 weeks later the buyer could not name the company. Breakout sent limited-edition Lego to target accounts instead, picked because Lego fans are passionate rather than because everyone likes it. The gift was paired with a LinkedIn and email follow-up asking whether it arrived, and 30% of recipients replied positively.2. A stunt is measurable when the input and the output are set before it runs.Breakout caps any stunt at $20K, above which another channel wins. The funeral, a hired hearse and a eulogy for cold outbound at Inbound, cost under $6K against a $90K to $100K booth. With a qualified lead costing $500 to $1,000 across channels, 6 qualified leads met the benchmark. Coffee vouchers of $5 to $10 for anyone who posted on the day made the result countable, and it returned 20 to 25 leads plus 8 to 10 inbound queries.3. Brand activations carry no pipeline target, by design.Gupta separates the decision to run a stunt from its direct pipeline contribution, and plans none against it. The targets are social: impressions, reach and activation, because stunts feed LinkedIn, the mailing list and later sales conversations. He needs a stunt to be a smashing success about 20% of the time to earn the money back. The cardboard-sign stunt, which drew an 80 to 90% open rate on its Polaroid follow-ups and no conversions, sits inside that tolerance.4. Creating attention and converting it are different businesses.Rick named the tension: Breakout sells software that automates the buyer’s first conversation while winning with things software cannot do. Gupta’s answer is that Breakout does not create attention, it converts it. 60% of its closed-won comes from inbound and another 20% from network, and every one of those buyers goes through the agent, because the site has no form. That is also why Breakout sells to mid-market companies with traffic, and not to early-stage companies without it.5. At peak intent, the cost of a wrong answer is the pipeline itself.Outbound agents can get 5 emails wrong out of thousands and lose little, because those recipients were never going to convert. An inbound agent talks to a buyer who already has intent. Breakout learned this when large customers pointed the agent at thousands of pages, including a 5-year-old acquisition announcement for a product since shut down. Onboarding now starts with curating what the agent reads, and everything stale stays out.Sachin GuptaLinkedIn: https://www.linkedin.com/in/sachingupta006/ Breakout: https://www.getbreakout.ai/Rick Koleta (Host)LinkedIn: https://www.linkedin.com/in/rickkoleta/Instagram: https://www.instagram.com/rickkoletaGTM VaultLinkedIn: https://www.linkedin.com/company/gtmvaultYouTube: https://www.youtube.com/@gtmvaultWebsite: https://gtmvault.co/The GTM Vault Podcast - subscribe on your platform of choiceSubscribing on YouTube, Apple, or Spotify is what tells each platform to put the next episode in front of more operators, which is how the show keeps booking the guests worth an hour of your week.Previous episodes includeSee all previous episodes hereSachin’s stack is on the record. See yours against it: 7 taps at pro.gtmvault.co/survey, and the State of the GTM Stack 2027 report ships in March, free.The GTM Vault Podcast is a breakdown of how high-growth companies design, test, and scale revenue architecture. Visit gtmvault.co for more episodes, the GTM Vault newsletter, and Show Me Your Stack. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.gtmvault.co/subscribe
GTM Vault is a breakdown of how high-growth companies design, test, and scale revenue architecture.Every dollar in revenue AI is chasing the agent that sells. Almost nobody is funding the one that keeps the system those agents run on from rotting, which is strange, because the org they all read from is the thing quietly decaying underneath them. Jelle van Geuns spent over a decade inside hundreds of Salesforce orgs before he built anything, and what he saw was not breakage. It was accumulation. The founder and CEO of Cirra AI joins GTM Vault for episode 54 to explain why the admin layer is a company rather than a feature, where the guardrails on an AI that can touch Setup actually sit, and why an independent survives the platform vendor and the frontier labs on grounding and history rather than on model capability.About Cirra AICirra AI builds an AI admin for enterprise software, starting with Salesforce: someone describes a change in plain language, and Cirra reads the org, proposes an implementation, builds the metadata in a sandbox, and documents what it did and why. It was founded in 2024 by Jelle van Geuns, a Dutch-born engineer and decade-plus veteran of the Salesforce ecosystem, and runs out of San Francisco. It ships what it calls the first commercial Model Context Protocol server built for Salesforce administration, so an admin can change an org from Claude or another agent rather than logging into Setup. Cirra has raised no outside capital; before it, Jelle bootstrapped Decisions on Demand, a lead-routing AppExchange app, to seven-figure ARR without external funding.Watch or listen now across YouTube, Apple Podcasts, and SpotifyEpisode highlights(0:00) The Agent Nobody Is Funding(3:30) Orgs Do Not Break, They Accumulate(5:01) The Backlog At A 200 Person Company(7:02) The Seven Figure ARR Correction And No Outside Capital(8:22) One Change Request From Slack To Production(10:27) Four Guardrails On An AI That Touches Setup(14:57) Why Agentforce Does Not Absorb This(16:42) What Stays Yours When The Models Improve(22:19) The Day The Platform Turns Hostile(25:47) The Autonomous AI Admin And What Early Autonomy Got WrongWhat you’ll learn:* Why orgs do not break but accumulate, and why only about 2% of surveyed Salesforce orgs are described as clean and well maintained* The norm at a 200-person company: one part-time admin, often the VP of sales, and 40 to 60 open requests that never shorten* What the backlog costs where no line item shows it: a spreadsheet on the side, a shadow Asana instance, and reports that stop matching reality* One change request end to end, from a Slack sentence to a confirmation gate, an admin approval, a sandbox build, and the customer’s own DevOps pipeline* Four guardrails on an AI that touches Setup: the OAuth permission ceiling, role-based access, read before write, and the promotion flow* Why Agentforce does not absorb the admin layer, and the tell: Salesforce’s own MCP servers read metadata but will not change it* The headless bet, which is configuring Salesforce from the agent you already use rather than logging into the platform to do it* What stays durable when the models improve: org-specific grounding, the accumulated record of why a change was made, and the approval trail* The answer to “why not wait for Salesforce”, which is 2 or 3 releases, 9 months to a year, and a year and a half before it hardens* The 18-month test: first-level admin traffic goes to the AI, and nothing gets reworked after it is approvedKey takeaways1. Orgs do not break. They accumulate, because nobody is accountable for removal.Every quarter somebody adds a field, a validation rule, a flow, or a permission change, and each one has a rationale at the time. Nothing comes back out, because removing it might break something and the reason it exists has been forgotten. After 5 years you have 3 fields that do the same thing and nobody knows which one is live, plus a flow nobody dares open. A Salesforce Ben survey Jelle cites puts orgs described as clean and well maintained at about 2%.2. The backlog is not an admin staffing problem. It is a system of record problem.The norm Jelle describes is 200 people with one part-time admin who grew into the role without choosing it, often revenue operations or the VP of sales, and a queue of 40, 60 or more open requests that never shortens because new ones arrive daily. Sales does not wait. They build a spreadsheet on the side, or stand up an Asana instance and run deal management there. The reports stop reflecting reality, because reality moved out of the CRM.3. The guardrail is not the model. It is the permission ceiling and the read-before-write rule.Cirra runs as a connected user through OAuth and can never do anything that user could not do by hand in Setup, so the blast radius is bounded before any model reasoning happens. Role-based access sits on top of that, fine-grained by individual, by role, even by object and field, so a team can start read-only in production and open it up incrementally. Then it reads before it writes and proposes a specific plan first, which is what stops it adding a fourth duplicate field. Promotion still runs through whatever the customer already has, DevOps Center, Flosum or Gearset.4. Against the platform the bet is headless. Against the labs the bet is the history.Agentforce builds agents that face customers and employees; changing how the org is built is a different job, and the tell is that Salesforce’s own MCP servers read and write records and read metadata but will not change metadata. So Cirra’s bet is that you configure Salesforce from the AI tools you already use rather than logging into the platform to do it. The frontier models know Salesforce in general but have no access to your specific org. What compounds is the record of what was there, what changed, and the reasoning behind it.5. Bootstrapping is a timing instrument, not a badge.To be precise, and Jelle corrected this on air, the seven-figure ARR belongs to Decisions on Demand, not to Cirra. The reason for no outside capital is that neither the models nor the standards around them were ready for a full AI admin when he started, so he iterated through product formats that did not work without burning a round to find out. That patience is also the sales answer to “why not wait for Salesforce”: 2 or 3 releases is 9 months to a year, and Salesforce has a history of taking a couple more releases to harden what it ships.Jelle van Geuns* LinkedIn: https://www.linkedin.com/in/jgeuns/* Cirra AI: https://cirra.aiRick Koleta (Host)* LinkedIn: https://www.linkedin.com/in/rickkoleta/* Newsletter: https://gtmvault.co/GTM Vault* LinkedIn: https://www.linkedin.com/company/gtmvault* YouTube: https://www.youtube.com/@gtmvault?sub_confirmation=1* Website: https://gtmvault.co/The GTM Vault Podcast - subscribe on your platform of choiceSubscribing on YouTube, Apple, or Spotify is what tells each platform to put the next episode in front of more operators, which is how the show keeps booking the guests worth an hour of your week.Previous episodes includeSee all previous episodes here: https://www.gtmvault.co/podcastThe GTM Vault Podcast is a breakdown of how high-growth companies design, test, and scale revenue architecture. Visit gtmvault.co for more episodes, the GTM Vault newsletter, Show Me Your Stack, and Coffee Walk, the on-location founder series that opens with Barbaros Ozbugutu in Istanbul. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.gtmvault.co/subscribe
GTM Vault is a breakdown of how high-growth companies design, test, and scale revenue architecture.This episode in the Library: the written playbook, the stack and the full transcript, free and open. Open it at pro.gtmvault.co.Every vendor in your inbox is selling a signal, and most teams respond by picking the best one and wiring it to a sequence. The data says that is the wrong question. Accounts carrying more than eight signals generate nineteen times the opportunities of accounts carrying one or two, while the best single signal in B2B, champions on the move, tops out at about twelve percent of opportunities on its own. Christian Kletzl built UserGems on that one signal, in a one-week build in a cabin in the Austrian Alps that sold its first contract for $50,000. The co-founder and CEO joins GTM Vault for episode 53 to explain why he has spent the years since moving past it: how stacked signals get scored into one daily ranking, why if-this-then-that routing burns your best contacts, and why he gave up the dashboard to get there.About UserGemsUserGems tracks when a company's past champions, users and buyers change jobs, then scores that against more than twenty other signal types to rank which accounts and people a revenue team should work on a given day. It was founded by the Austrian twin brothers Christian Kletzl, the CEO, and Stephan Kletzl, the CTO, and runs out of San Francisco, having started as SmartHires, their Y Combinator W15 company. Its customers are mid-market and enterprise B2B revenue teams, including Procore, Medallia, UserTesting and Sisense. UserGems has raised $22.4M, including a $20M Series A led by Craft Ventures with Battery Ventures, Tiger Global and Uncork Capital.Watch or listen now across YouTube, Apple Podcasts, and SpotifyEpisode highlights(0:00) Cold Open: Eight Signals, Nineteen Times the Opportunities(4:01) A Cabin in the Austrian Alps and a One-Week Build(5:38) Why Champions on the Move Tops Out at Twelve Percent(10:53) The If-This-Then-That Mistake and What It Costs You(13:25) Stacking Signals Into One Daily Ranking of Fifteen(16:34) Twenty Million Accounts, Six Hundred Attributes, One ICP Score(21:52) Where Gem-E Stops and the Rep Starts(26:34) Going Headless Over MCP and Who Owns the Outcome(34:41) Data Accuracy and the Ten Percent Trust Line(37:14) Publishing Pricing and the Case for Self-Selecting OutWhat you’ll learn:* The finding that accounts carrying more than eight signals produce nineteen times the opportunities of accounts carrying one or two* Why champions on the move generates about twelve percent of opportunities, why that share holds across organizations, and why it compounds on its own* Why if-this-then-that routing spends your best information on whichever signal fires first, and burns the account for six months doing it* How twenty million accounts and six hundred attributes become one ICP ranking, built by comparing your customers against everyone who never bought* The two exceptions that should still fire instantly, and why everything else belongs in a daily ranked list of fifteen* Why a ten percent error rate is where reps stop trusting a signal motion whatever its coverage, and the case against waterfall enrichment* How UserGems doubled SDR-generated opportunities from nine to nineteen a month, hired more SDRs on the back of it, and kept calls human* Which signals a competitor can buy, which ones only you can create, and why that line decides durability* What changed in AI-written outbound around September 2025, and how the customer complaints flipped* Why publishing a $40K entry price is a qualification mechanism rather than a pricing decision, and the math Kletzl expects a buyer to runKey takeaways1. The best signal in B2B has a ceiling, and the ceiling is the whole argument.Champions on the move generates about twelve percent of opportunities, and Kletzl says that share holds across organizations. It compounds on its own, because a larger customer base produces more job changes. The mechanism underneath is durable: roughly eighty percent of people stay in the same industry and eighty percent stay at the same company size, so a champion carries the same problems into the next role. But twelve percent is not what a CMO wakes up thinking about, which left UserGems valuable and peripheral at once. Every product decision since has been an attempt to reach fifty.2. If-this-then-that routing spends your best information on your first signal.Kletzl names this as the most common mistake he sees, and says the easier the tooling makes it, the more it happens. A signal fires, a sequence launches, and that person is now off limits for six months. Better information arriving next week has nowhere to go. Two exceptions still fire instantly, a past champion and a pricing or comparison page visit. Everything else waits for the ranking.3. Stacking signals beats picking signals, by a factor of nineteen.Accounts carrying more than eight individual signals generate nineteen times the opportunities of accounts carrying one or two. That reframes the question most teams bring him. They want to know which signals to switch on, when the compounding sits in how many land on the same account and how they are scored together. It is also why he refuses to start a new customer on workflows, and starts on ICP definition instead. Kletzl will not name the signals behind his best customers’ motions, because at the top end it is no longer about any individual one.4. Data accuracy is the foundation, and the tolerance is ten percent.Above a ten percent error rate, Kletzl says people stop trusting the motion, and no amount of coverage recovers it. That applies to every signal and not just job changes, because a website visitor with a stale title is a dead signal the same way a nine-month-old job change is. It is also his case against waterfall enrichment, which buys quantity at the cost of knowing whether any of it is right. UserGems would rather return nothing than return something wrong.5. Going headless gives up the interface and keeps the outcome.UserGems exposed the whole platform over MCP, including campaign setup, audience building, list uploads and persona creation, not just data export. Kletzl puts that in the middle of three phases: humans clicking through a browser, then humans directing AI, then AI tuning campaigns against its own results. He is comfortable losing the dashboard because the end user was never in it. The buyer was always buying opportunities generated, and ownership of the outcome does not move when ownership of the interface does.Christian KletzlLinkedIn: https://www.linkedin.com/in/christiankletzl/ UserGems: https://www.usergems.com/Rick Koleta (Host)LinkedIn: https://www.linkedin.com/in/rickkoleta/ Newsletter: https://gtmvault.co/GTM VaultLinkedIn: https://www.linkedin.com/company/gtmvault YouTube: https://www.youtube.com/@gtmvault?sub_confirmation=1Website: https://gtmvault.co/The GTM Vault Podcast - subscribe on your platform of choiceSubscribing on YouTube, Apple, or Spotify is what tells each platform to put the next episode in front of more operators, which is how the show keeps booking the guests worth an hour of your week.Previous episodes includeSee all previous episodes here: https://www.gtmvault.co/podcastThe GTM Vault Podcast is a breakdown of how high-growth companies design, test, and scale revenue architecture. Visit gtmvault.co for more episodes, the GTM Vault newsletter, Show Me Your Stack, and Coffee Walk, the on-location founder series that opens with Barbaros Ozbugutu in Istanbul.Full transcript, searchable to the timestamp: pro.gtmvault.co/episodes/gtm-53. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.gtmvault.co/subscribe
GTM Vault is a breakdown of how high-growth companies design, test, and scale revenue architecture. 27,000+ operators read it. Join them.This episode in the Library: the written playbook, the stack and the full transcript, free and open. Open it at pro.gtmvault.co.Every vendor in revenue AI is pointed the same way. Automate the seller, draft the outbound, hand the follow-up to an agent. Which leaves a question nobody on that side of the market is funded to answer: what happens to the conversations an agent does not get to take, and who is responsible for making the humans in them better. Ariel Hitron has been running the other way since 2019, building AI that plays the buyer so the seller gets better. The co-founder and CEO of Second Nature joins GTM Vault for episode 52 to explain why an enterprise buyer signing off on a large number is not buying capability, where he draws the line between an agent conversation and a human one, and why the training product was only ever the wedge.About Second NatureSecond Nature builds a data model of a company’s sales playbook out of its recorded calls, collateral, scripts and CRM stage criteria, then generates AI roleplays that let reps practice against it and scores them afterwards against that company’s own criteria, in more than twenty languages. It was founded in 2019 by Ariel Hitron, who helped scale Kaltura from startup to global business, and Alon Shalita, a former lead engineer at Facebook, and runs out of New York. Its customers include Zoom, Oracle, Adobe, Teleperformance and Check Point, and the company reports sales up more than twenty percent after an average of thirty minutes of practice per trainee, with onboarding at some accounts cut by three weeks off a nine week process. Second Nature has raised $38M in total, including a $22M Series B in October 2025 led by Sienna VC with Bright Pixel, StageOne Ventures, Cardumen, Signals VC and Zoom, which is also a customer.Watch or listen now across YouTube, Apple Podcasts, and SpotifyDownload the transcriptEpisode highlights(0:00) Who Do You Hold Accountable When It Breaks(1:38) The Category Automates the Seller. He Went the Other Way(5:07) 2019, Before ChatGPT, and the Objection Every Enterprise Gave(7:33) Digitizing the Playbook Into One Machine-Readable Model(9:35) What the Scoring Layer Catches That a Manager Misses(13:22) Which Conversations Should Go to an Agent, and Which Never(19:18) Bundling, Unbundling, and Why Enterprises Unbundle Upward(24:53) From Roleplay to Agentic Sales(30:49) Demos Are Free. Maintenance Is the Moat(38:24) The 2028 Revenue Org and the Test That Would Prove Him WrongWhat you’ll learn:* Why an enterprise buyer approving fifty thousand or five hundred thousand dollars is buying accountability, which an agent has none of* The two variables that draw the human line: size of the investment and how far the product sits from a commodity* Why personality and preference is a third factor he names, and pointedly does not rank against the other two* A playbook lives in four places at once, and none of them can be trained against until it is compiled into one model* The machine does not beat a sales manager on judgment, it beats them on patience across a seven-parameter scorecard* Building a hostile buyer is trivial, and the silent one is the hardest to sit across from* Why enterprises are unbundling their enablement suite upward, buying the one capability the platform layer has not absorbed* Demos are free and maintenance is the moat: what happens to every in-house roleplay build after the hackathon* Why large competitors shipping roleplay as a feature made the sale easier rather than harder, and what he had to stop arguing* The falsifiable test, on a clock: if the sales workforce shrinks, he was wrongKey takeaways1. Enterprise buyers are not buying capability. They are buying someone to answer the phone when it breaks.Ariel gives away the whole capability argument without a fight. An agent can run discovery, ask the questions, solution, and put a proposal in front of you. What it cannot do is absorb the consequence, and a buyer signing off on fifty thousand or five hundred thousand dollars has put their own standing inside the company on the line to do it.2. Two variables draw the human line, and a third one he names without ranking.Size of the investment, and how far the product sits from a commodity. He puts the first threshold somewhere between ten and fifty thousand dollars, and the second at the gap between restocking something you have bought before and deploying a system nobody in the market can vouch for yet. Toilet paper goes to a bot and a new ERP does not, and the third factor he names is personality and preference, which he leaves unranked against the other two.3. A playbook you have not compiled cannot be trained against, by an agent or a new hire.The playbook is real and it is scattered across four surfaces: product marketing decks, tribal knowledge sitting in reps’ heads, recorded live calls, and the stage entry and exit criteria in the CRM. Second Nature’s job is to consolidate all of it into one machine-readable model of how that specific company sells. Every org runs its own mutation of MEDDPICC or SPIN, so the named methodology tells you almost nothing about the process that survives contact.4. The scoring layer does not beat a manager on judgment. It beats them on patience.Humans read the room, catch the nuance, and bring everything they have seen before into the call. What they do not do is walk a seven-parameter scorecard, call by call, rep by rep, and justify each score. Ariel’s framing is that sales leaders are not accountants, so the feedback that lands is that you were not convincing, not exciting, not bringing energy, which is true and unusable, and the machine advantage is consistency rather than insight.5. Demos are free. Maintenance is the moat, and it is why the in-house build decays.The pattern repeats across enterprises: a hackathon, an internal team, a working prototype, then decay. The builders return to their day jobs, the voices go stale, credits run out, access controls are wrong, and nothing consolidates the data. Every layer underneath moves fast enough, from the models to speech to text to the agent harnesses, that maintaining it becomes irrational for anyone whose job is not this.Ariel Hitron* LinkedIn: https://www.linkedin.com/in/arielhitron/* Second Nature: https://www.secondnature.aiRick Koleta (Host)* LinkedIn: https://www.linkedin.com/in/rickkoleta/* Newsletter: https://gtmvault.co/GTM Vault* LinkedIn: https://www.linkedin.com/company/gtmvault* YouTube: https://www.youtube.com/@gtmvault?sub_confirmation=1* Website: https://gtmvault.co/Previous episodes includeSee all previous episodes here.The GTM Vault Podcast is a breakdown of how high-growth companies design, test, and scale revenue architecture. Visit gtmvault.co for more episodes, the GTM Vault newsletter, and Show Me Your Stack.Full transcript, searchable to the timestamp: pro.gtmvault.co/episodes/gtm-52. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.gtmvault.co/subscribe
GTM Vault is a breakdown of how high-growth companies design, test, and scale revenue architecture. 27,000+ operators read it. Join them.This episode in the Library: the written playbook, the stack and the full transcript, free and open. Open it at pro.gtmvault.co.Your CRM knows the deal, the stage, and the contact's title. It does not know which person on your team can walk you into the account, or that the champion who changed jobs last quarter is now a warm door somewhere else. Olivier Roth built a company on that gap. The co-founder and Chief Growth Officer of The Swarm joins GTM Vault for episode 51 to explain why the relationship layer sits alongside the CRM rather than inside it, why it travels to wherever the work happens, and who ends up owning the relationship when it does.About The SwarmThe Swarm turns a company’s extended network into a data layer its sales team can query. Founded in San Francisco in 2021 by David Connors, Michal Bil, and Olivier Roth, it grew out of Connors’ previous company, Automately.io, which Sequoia acquired, and the tools he then built so Sequoia’s portfolio could tap the firm’s network. The graph now covers 580M+ people and 100M+ company profiles, pipes into HubSpot, Attio, Clay, and any agent via MCP, and in 2026 the company acquired Commsor, the competitor that coined the go-to-network category. The Swarm has raised $8M in total, including a $4M seed led by 500 Global and a round backed by HubSpot Ventures, Motivate Ventures, and TRAC VC.Watch or listen now across YouTube, Apple Podcasts, and SpotifyDownload the transcriptEpisode highlights(0:00) LinkedIn Is Your Audience, Not Your Network(2:44) What A CRM Cannot Capture, And The Anti-CRM(4:07) The Six Edges Of A Relationship Graph(7:32) Where The Graph Lives: Inside, Beside, Or Under The CRM(12:40) One Motion: LinkedIn Profile To Claude Prompt To Clay Workflow(15:06) 500 Million Profiles, Freshness, And The Consent Architecture(19:05) Potential, Likely, Confirmed: How The Graph Scores A Path(23:39) The Commsor Acquisition And The Size Of The Category(26:32) Swarm Send, 40 Percent Of Pipeline, And Three Intros Every Monday(32:36) Rapid FireWhat you’ll learn:* Why the CRM is a system of the past, and why the deals that could exist never appear in it* LinkedIn is your audience, not your network: why a binary connection tells you nothing and most people know five to ten percent of theirs* The six edges in a relationship graph: work overlaps, education overlaps, investor overlaps, LinkedIn connections, LinkedIn engagement, and email and calendar contact* Potential, likely, confirmed: the scoring that says how sure the graph is that two people know each other* Why the relationship layer is a building block that travels, and why the CRM stays the central nervous system anyway* How a rep uses the graph from a LinkedIn profile, from a Claude prompt, and from a stalled-deals query inside HubSpot* Freshness is what breaks at scale, and why The Swarm built its own 500-million-profile data layer instead of buying one* The consent architecture: what is passive, what is user-permissioned, and why every company’s swarm is siloed* Why orchestration, not data, is the next thing to build, and what three intros every Monday morning looks like in practice* The test that would prove the bet: thirty percent of pipeline from warm intros, visible in the CRM’s own reportsKey takeaways* The CRM records the past. The deals that could exist live somewhere it cannot see.A CRM records interactions that already happened, tied to deals that already exist, and that is its job. It holds nothing on the other side: every potential deal reachable through someone attached to the company. The Swarm pools the networks of a company's investors, advisors, partners, employees and best customers, which Olivier calls an anti-CRM, peering forward at what the relationships make possible rather than backward at what closed.* LinkedIn is your audience, not your network, and a connection alone scores as "potential."A LinkedIn connection is binary, no context, and Olivier puts the share of connections people know at five to ten percent. So The Swarm stacks six edges instead: work overlaps, education overlaps, investor overlaps, LinkedIn connections, LinkedIn engagement, and email and calendar contact. Each relationship scores potential, likely, or confirmed: emailed, met and connected is confirmed; a single meeting or a short work overlap is likely; a LinkedIn connection and nothing else is potential.* The layer is a building block, and the CRM stays the central nervous system.Go-to-market engineers build agents on The Swarm's API; sellers use it inside HubSpot, Attio and Clarify, with Salesforce in progress, and in Clay workflows. With the HubSpot MCP a rep can take every deal stalled for ninety days and ask for a champion or investor in the swarm who can unstick it. Olivier still calls the CRM the central nervous system, because that is where revenue is recorded and reported, and the bet is that when every CRM ships a warm intro tab, the tab is built on The Swarm.* Freshness is what breaks at scale, so they built the data layer themselves, with the consent lines drawn in advance.No existing provider was fresh enough, so The Swarm built its own: roughly 500 million profiles, 100 million companies, job-change tracking, and a live enrichment endpoint that refreshes a profile on demand. Work, education and investor overlaps are computed passively from public history; LinkedIn connections, email and calendar are user-permissioned and visible only to the admins of that company's swarm. Every company's data sits in its own silo.* The data took five years. Orchestration is what comes next, and the proof is forty percent of their own pipeline.Olivier vibe-coded Swarm Send, which ranks the best-scored intro paths to target accounts and drafts the requests; roughly forty percent of The Swarm's pipeline now comes through intros. The pattern he wants to productize is one a customer's go-to-market engineer already built: three intros dripped to each seller every Monday morning. The test that proves the bet is a report, not a case study: thirty percent of pipeline from warm intros, visible as fields inside the CRM, next to paid ads on the same chart.The LibraryEvery episode of this show, and every Show Me Your Stack, now lives as a written playbook in one searchable place. Fifty podcast episodes, eight stack breakdowns, and the tools behind each one indexed across all of them, so you can pull every episode where Clay or Claude Code shows up before you wire it yourself. It is part of Pro. If you are a member, sign in at pro.gtmvault.co with your subscriber email. If you are not, this is the reason to upgrade.Olivier Roth* LinkedIn: https://www.linkedin.com/in/olivier-roth/* The Swarm: https://www.theswarm.comRick Koleta (Host)* LinkedIn: https://www.linkedin.com/in/rickkoleta/* Newsletter: https://gtmvault.co/GTM Vault* LinkedIn: https://www.linkedin.com/company/gtmvault* YouTube: https://www.youtube.com/@gtmvault* Website: https://gtmvault.co/Previous episodes includeSee all previous episodes hereThe GTM Vault Podcast is a breakdown of how high-growth companies design, test, and scale revenue architecture. Visit gtmvault.co for more episodes, the GTM Vault newsletter, and Show Me Your Stack.Every tool operators name on GTM Vault is in The Stack Index, placed on the layer of revenue architecture it serves. Build your own stack at pro.gtmvault.co/stack.Full transcript, searchable to the timestamp: pro.gtmvault.co/episodes/gtm-51.On the record from this episode: Lovable. The link shows every operator who runs it on The Stack Index. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.gtmvault.co/subscribe








