All podcasts / Lenny's Podcast / Summary

We replaced our sales team with 20 AI agents—here's what happened | Jason Lemkin (SaaStr)

2026-01-01 - 102 min - source - Read full transcript

Lenny Rachitsky (host)Jason Lemkin

Key insights

SaaStr replaced a roughly 10-person GTM team with 1.2 humans plus 20 AI agents at equivalent business output.

Lemkin describes literally relabeling the sales floor's desks with agent names (Artie for Artisan, Quali for Qualified, Reply for Replit) after two paid SDRs quit on-site at SaaStr's annual event. Net revenue productivity came out roughly the same as with 10 humans, but scales because it runs on software rather than headcount.

ai-sales-transformation

The classic cadence-based SDR and inbound-lead-qualifying BDR roles will be mostly extinct within 12 months.

Lemkin argues there is no reason to keep a 21-year-old sending scripted outbound emails or making prospects wait a day to be qualified when an AI agent can do both instantly, around the clock, and can already fully qualify website visitors without them realizing it.

ai-sales-transformation

Account executives are safer near-term but the job is shrinking: roughly 70% of AE jobs are safe through the next year, declining toward 40-50% after that.

He sees no structural reason a well-trained agent can't close deals with limited price negotiation, especially for buyers who would rather interact with a smart AI than a mediocre human. Complex, high-touch enterprise sales remain the holdout.

ai-sales-transformation

AI GTM agents don't work out of the box - they require a deliberate, weeks-long training loop, and vendors who claim otherwise are lying.

The real process: give the agent your website, wiki, and training docs to ingest, let it generate clarifying questions, answer them, then spend roughly an hour a day for about 30 days correcting its mistakes (including hallucinations) until it performs like your best rep. Companies that skip this step and hand an untrained agent to junior staff get zero ROI.

agent-orchestration

Managing a fleet of AI agents is a demanding, near full-time job, not a set-and-forget deployment.

SaaStr's chief AI officer, Amelia, spends 10-15 hours a week reviewing agent output because the agents work nights, weekends, and holidays and never stop generating things that need human review. Lemkin says SaaStr may already be at capacity and doesn't know when it will deploy a 21st agent.

agent-orchestration

When choosing an AI GTM vendor, weight the forward-deployed-engineer relationship as heavily as the product itself.

Lemkin's first two picks (Artisan, Qualified) succeeded specifically because those vendors were the only ones willing to help with no upfront guarantee, while a bigger competitor demanded $100K up front and another declined outright, fearing bad PR if it failed publicly with SaaStr.

agent-orchestration

Most companies wildly underestimate how much untouched lead and traffic data they already have to train an agent on.

Lemkin pushes back on founders who say they're 'too small' for AI GTM tools - a company with 300 customers often has tens of thousands of website visitors or CRM records that zero humans have ever contacted, which is more than enough scale for an agent to work with.

gtm-strategy-2026

2026 GTM is defined by 'everyone in market at once' - in-market rates for AI tools now exceed 50% of prospects in some categories versus a historical 3-5% baseline.

This is reshaping outbound economics on both ends: hyper-growth AI companies can't service the inbound demand they already have, while legacy SaaS companies face buyers with unprecedented urgency but also unprecedented scrutiny, since ROI now has to be proven before the contract is signed.

gtm-strategy-2026

The GTM plays still work, but the pre-AI playbooks built around them don't produce enough ROI anymore.

Outbound, webinars, podcasts, and events all still convert, according to Lemkin, but growth has decelerated so much for companies running old playbooks that it looks like nothing works. Fast-growing AI companies run the same plays with a PLG-first, selective-response posture instead.

gtm-strategy-2026

Well-trained AI-written sales emails already beat the average human rep's emails because most human outbound is mediocre.

Lemkin recalls inheriting a team of reps from Adobe whose emails were the worst he'd ever read. He argues the bar for 'good enough' AI GTM copy isn't as high as people assume - training an agent on your best rep's template and letting it A/B test variants reliably outperforms a mid-pack human, and recipients largely don't care that it's AI-written as long as it adds value and responds instantly.

ai-sales-transformation

The single highest-leverage career move for a salesperson worried about AI is to personally deploy and train an agent themselves.

Lemkin's concrete advice: pick any leading vendor, do the full ingestion-and-training cycle yourself over about 30 days and 50-60 hours, and you become 'hyper employable' - most companies asking for GTM hires can't find anyone who has actually done this end to end.

future-of-work

Don't quit a company with happy, paying customers to chase a new AI startup - the business you're already at is probably your best bet.

Lemkin cites founders and executives leaving nine-figure equity positions or profitable customer bases to chase the next hot AI idea, arguing it's much harder to rebuild 500 happy customers from scratch than to turn your current company into your own AI-native business.

future-of-work

Media referenced

Pluribus - show - Apple TV show Lemkin uses as a metaphor for where AI GTM is heading - agents sharing a hive mind of customer data across a company, working 24/7
Everything That Breaks on the Way to One Billion (SaaStr talk with Ben Chestnut) - other - Lemkin's favorite SaaStr talk of all time - Mailchimp's co-founder on the deal almost falling apart post-acquisition and challenging conventional B2B metrics
Rippling's Secrets to Hyper Growth (SaaStr talk with Matt Plank and Sam Blond) - other - Recommended GTM-focused talk from two CROs (Rippling's and ex-Brex's) with an honest, learner's-mindset take on the revenue playbook

Companies

SaaStr - Lemkin's B2B founder community and events business; the company whose sales org he rebuilt around AI agents
Delphi - Digital-clone platform Lemkin used to build his own AI persona ('digital Jason'), which unexpectedly closed a $70K sponsorship on its own and sparked the AI-GTM push
Artisan - YC-backed outbound AI SDR vendor SaaStr picked first because it was the only vendor willing to help with no upfront guarantees; grew from near-zero to $10M in revenue this year
Qualified - Inbound-qualification AI vendor, founded by Salesforce's ex-CMO, that SaaStr used to instantly respond to and qualify inbound sponsorship/ticket leads
Salesforce / Agentforce - Used for reactivating stale leads that human reps deprioritized; achieved a 70% response rate; Salesforce itself is becoming the shared database ('hive mind') for AI GTM agents
Replit - No-code app builder Lemkin uses obsessively (self-described top 1% user) to build internal tools like a valuation calculator and pitch-deck reviewer
Vercel - Cited repeatedly as an example of a hyper-growth AI company running lean go-to-market teams by design, even as they scale toward enterprise
Owner.com - AI-for-restaurants company Lemkin advises; pushing reps toward $3-5M in revenue per rep with AI support versus $300-500K a few years ago
Momentum / Attention - RevOps tools that automatically log every rep action into the CRM in real time, creating full transparency that exposed at least one underperforming rep who quit the day it rolled out

Techniques and frameworks

Forward-deployed engineer (FDE) model - Vendor-side technical staff whose job is to make sure the customer's AI agent actually works on go-live day, inverting the classic SE-to-rep ratio; borrowed from Palantir's playbook
Agent training loop: ingest, train, correct, iterate - The actual mechanics of deploying a GTM agent - upload source data, answer the agent's clarifying questions, correct daily mistakes for about 30 days until it performs like your best rep
Incognito-mode customer journey audit - Lemkin's recommended quarterly exercise: go through your own support, sales, and signup flow anonymously to find the parts bad enough to make you 'cry,' then fix that one with an agent

Summary

Jason Lemkin returns to Lenny's Podcast roughly eighteen months after their first conversation to describe how he tore down and rebuilt SaaStr's entire go-to-market team around AI agents. What used to be eight or nine full-time SDRs, BDRs, and AEs is now 1.2 humans (one full-time AE plus 20% of chief AI officer Amelia's time) and 20 AI agents, producing roughly the same revenue. The turning point was almost accidental: after two well-paid sales reps quit on the spot at SaaStr's annual conference, Lemkin decided he was done hiring humans for jobs he'd already watched fail multiple times, especially after a general-purpose "digital Jason" agent built on Delphi had unexpectedly closed a $70K sponsorship deal on its own.

The bulk of the conversation is a practical field guide to what actually works. Lemkin walks through how SaaStr picked vendors (Artisan for outbound, Qualified for inbound, Agentforce for lead reactivation) based less on features and more on which vendors were willing to actually help train the agents - a distinction he treats as decisive, since untrained AI GTM software from any vendor underperforms trained software from a competitor. He's blunt that deploying these tools is real work: you ingest your data, answer the agent's clarifying questions, and spend roughly an hour a day for about 30 days correcting its mistakes before it starts to resemble your best salesperson. Skipping that step, which he says most panicked enterprises try to do, guarantees failure.

Lemkin's read on the sales profession is sharply bifurcated. Cadence-based SDRs and inbound-qualifying BDRs are, in his estimation, mostly extinct within a year - there's no defensible reason to keep a human sending scripted emails or making prospects wait a day to be qualified. AEs are safer for now (he estimates 70% of jobs intact through the next year, declining toward 40-50% after) because complex negotiation and enterprise trust-building still favor humans, though he's skeptical that "being a people person" is a durable moat when AI can already run a warm, responsive conversation at scale. His clearest advice for anyone worried about their job: personally deploy and train an agent yourself rather than waiting for your company or an agency to do it - having done that once, end to end, makes you dramatically more employable in the roles opening up.

Zooming out, Lemkin frames 2026 as a market where "the plays still work, but the playbooks are broken." Outbound, events, and webinars all still convert, but pre-AI playbooks no longer generate enough ROI given how much more competitive and urgent the buying environment has become - he cites in-market rates in some categories now exceeding 50% of prospects, versus a historical 3-5% baseline. That dynamic cuts both ways: hyper-growth companies like Vercel and Replit can't service the inbound demand they already have, while legacy SaaS companies face buyers with far higher expectations for proof of ROI before signing. He closes with a recurring theme: this is the best and busiest time to be building in software, and the advice for founders and reps alike is the same - lean in, train the tools yourself, and don't quit a company with happy customers to chase the next hot idea.

The episode ends with a lightning round covering Lemkin's favorite SaaStr talks (Ben Chestnut on Mailchimp's near-collapse pre-acquisition, and Rippling's Sam Blond and Matt Plank on the revenue playbook), his current obsession with the Apple TV show Pluribus as a metaphor for agents sharing a "hive mind" of customer data, and a closing piece of career advice: when in doubt, don't leave a job with happy customers.

Notable Quotes

"We're done with hiring humans and sales. We're done." - Jason Lemkin

"The classic SDR junior kid that is hired out of college to send emails... we don't need them... They should be extinct next year." - Jason Lemkin

"Agents work all night and they work weekends and they work on Christmas." - Jason Lemkin

"People have weaker relationships with their customers than they think." - Jason Lemkin

"The best startup you're ever going to have probably is the one you're working at today. Don't quit." - Jason Lemkin