Lenny's PodcastJason Lemkin: How 1.2 humans and 20 AI agents replaced sales
How outbound, inbound, and reactivation agents matched a 10-person sales team; just 1.2 humans replaced sales, displacing entry-level SDRs and BDRs first.
CHAPTERS
- 0:00 – 5:38
SaaStr’s “1.2 humans + 20 agents” sales org: the bold shift
Lenny opens with the headline: SaaStr went from ~10 go-to-market humans to 1 full-time AE, a part-time Chief AI Officer, and ~20 AI agents. Jason explains why he made the switch, what the office looks like now, and the core claim: similar output, far more efficiency and scalability.
- •Old GTM desks replaced by agent “seats” with names (Repli, Quali, Arti, etc.)
- •Agents run 24/7, including weekends and holidays
- •Motivation: reduce dependence on mid-performing hires and churn
- •Result so far: productivity roughly the same, but more efficient and scalable
- •AI replaces work people avoid and displaces “mid-pack and mediocre” performers
- 5:38 – 11:53
What SaaStr actually sells: sponsorships vs. tickets (and why it matters for agents)
Jason clarifies SaaStr’s business model and why the GTM work naturally splits into different motions. Sponsorship sales are higher ACV and more consultative; ticket sales are high-volume and workflow-heavy—perfect for automation.
- •SaaStr is a B2B founder community with events, media, and sponsorship revenue
- •Two core products: event sponsorships (~$70–80K avg) and tickets (hundreds to ~$2K)
- •Different motions require different agent workflows
- •Reactivating lapsed attendees/sponsors is a separate motion with its own agent
- •High-volume outreach becomes feasible when agents can touch the long tail
- 11:53 – 14:17
From 2–3 SDRs + up to 5 AEs to 1 AE: what changed and what didn’t
They quantify the before/after org design and performance expectations. Jason emphasizes he’d still hire great humans, but won’t accept slow ramp and weak product understanding when agents can execute consistently.
- •Previous org: 2–3 SDRs and up to 5 AEs supporting inbound/renewals/outbound
- •Current org: one AE + Amelia (20% time) + 20 agents handling top-of-funnel
- •Human “greats” remain valuable; average reps are increasingly replaceable
- •Key failing of traditional hiring: poor product mastery even after months
- •Agents don’t solve everything, but they eliminate many low-ROI human tasks
- 14:17 – 19:20
Go-to-market in the AI era: plays still work, but playbooks are breaking
Jason argues that outbound, events, webinars, and podcasts still work—but ROI and execution expectations have changed. The market is bifurcating between AI winners with massive demand and everyone else facing deceleration and efficiency pressure.
- •Classic plays (outbound, events, podcasts) still work; ROI is what shifted
- •Bifurcation: hypergrowth AI companies vs. slow/no-growth legacy SaaS
- •AI leaders face lead overflow; slower companies need ruthless efficiency
- •Example: fast growers sometimes just triage which leads to respond to
- •The environment rewards speed-to-response and operational excellence
- 19:20 – 23:44
The near-term future of SDRs/BDRs/AEs: what gets automated first
Jason predicts the biggest disruption hits email-cadence SDRs and inbound lead qualifiers first, because agents can respond instantly and accurately. AEs are safer near-term, but he expects gradual displacement as agents get better at closing.
- •Email-based SDR cadences: predicted ~90% displacement soon
- •Inbound qualification/BDR work: should largely vanish due to AI speed + quality
- •AE work: more protected today, but may shrink over time as agents improve
- •Entry-level career ladder becomes a broader societal/workforce challenge
- •Best reps gain leverage; mediocre reps lose ground
- 23:44 – 28:41
How to be in the “20% who thrive”: deploy an agent yourself (not theory)
Jason’s career advice is practical: pick one painful problem, choose a leading vendor, and personally do the training, QA, and iteration. Anyone who can put an agent into production becomes highly employable—because most teams still haven’t done the hands-on work.
- •Pick one use case (support, inbound qualification, SDR) and implement it end-to-end
- •Ingestion/training/orchestration are less scary than they sound—mostly structured setup + iteration
- •Success requires daily QA and corrections over ~30 days
- •Many exec teams still think tools work “out of the box” (they don’t)
- •Doing it yourself creates repeatable capability; second deployment is dramatically easier
- 28:41 – 30:04
Don’t build your own AI GTM stack: why buy beats build (for most companies)
Jason draws a clear line between fun internal tool-building and mission-critical GTM automation. He argues most companies should buy because the products are good enough, the market evolves too fast, and maintenance/obsolescence risk is high.
- •GTM agents are not worth building internally unless you’re uniquely equipped (e.g., Vercel-style talent)
- •Cost of tools vs. engineering + maintenance favors buying for most orgs
- •Innovation pace makes internal versions obsolete quickly
- •Jason builds many apps for fun, but none of SaaStr’s GTM automation is homegrown
- •Core principle: treat agent platforms like Notion—possible to build, rarely worth it
- 30:04 – 33:07
A tour of SaaStr’s agents: digital clone → outbound → inbound → reactivation
Jason walks through SaaStr’s sequence of deployments and what each agent does. The journey starts with a Delphi “digital clone” that evolved from Q&A into support and even selling, then expands into outbound email agents, inbound website qualification, and reactivation workflows.
- •Delphi started as ‘Digital Jason’ but became support and even closed a $70K sponsorship
- •Artisan used for outbound campaigns and scaling email volume
- •Qualified used for inbound website chat: qualification + meeting setting 24/7
- •Salesforce Agentforce used for reactivation and leads humans ignored
- •Different agents for different segments: high-end, low-end, lapsed
- 33:07 – 40:15
The hidden make-or-break factor: vendor help and forward-deployed engineers
Jason argues tool selection is less about feature matrices and more about whether the vendor will actively help you get to production. Because training is real work, the best product without hands-on enablement often underperforms a “good enough” product with great deployment support.
- •Early success came from vendors who ‘did the work’ alongside SaaStr
- •Add a new evaluation column: quality of FDE/solutions architect support
- •Many AI GTM tools are similar under the hood; deployment support is the differentiator
- •Some vendors refuse risky/high-visibility customers or demand big upfront fees
- •Best vendors sometimes turn away customers if they can’t ensure success
- 40:15 – 42:10
Operational reality: agent oversight, segmentation, and the “too many agents” ceiling
After scaling to ~20 agents, SaaStr discovered a new bottleneck: human oversight and conflict management. Agents produce nonstop output, requiring structured QA, segmentation, and ongoing iteration—creating an ‘orchestrator’ workload that can cap how many agents you can effectively run.
- •Agents require continuous review; they don’t become fully “set and forget”
- •Orchestration role is demanding because agents work 24/7
- •Multiple agents need segmentation to avoid overlapping outreach and conflicts
- •SaaStr may be at capacity for adding more agents due to oversight burden
- •Broader prediction: enterprises may hit ‘app/agent fatigue’ after too much change
- 42:10 – 47:32
How to make AI outbound emails actually good (and why humans aren’t the gold standard)
They tackle the biggest fear: AI spam. Jason explains that quality comes from training on your best human copy, then letting the agent A/B test and personalize using real data sources; he also notes many human sales emails are already poor, so “good enough” AI is often an upgrade.
- •Use your best rep/marketer’s copy as the base template; don’t start from generic AI writing
- •Modern model improvements (e.g., Claude-level capability) changed outcomes vs. 2024
- •Agents excel at variants and A/B testing; personalization is driven by CRM/website signals
- •Disclosure vs. hiding ‘it’s AI’ matters less than value + fast responsiveness
- •AI often outperforms mediocre human outreach because baseline human quality is low
- 47:32 – 53:48
Where humans still win: top-logo enterprise, in-person selling, and high-stakes nuance
Jason draws boundaries around what AI can’t reliably replace today: deeply bespoke, high-value account work and in-person sales dynamics. He frames AI as the solution for volume and neglected leads, while elite sellers focus on the few deals where human craft remains decisive.
- •For a small list of top target accounts, bespoke human work still dominates
- •AI shines when you have 5,000+ contacts/leads and need scalable touch
- •In-person/door-knocking and certain field sales motions remain unclear for AI replacement
- •Humans still matter most when negotiation, complexity, and relationship depth are high
- •AI’s core advantage is closing the gaps humans ignore (speed + coverage)
- 53:48 – 1:08:09
The ‘Chief Orchestrator’ role: who should run agents (and why it’s rarely Sales)
They zoom in on the oversight function (Amelia) and what it really requires. Jason suggests the best operators come from data-driven marketing, RevOps, product, or “nerdy” profiles who enjoy analytics and workflow design; traditional sales backgrounds alone usually aren’t enough.
- •Agent success requires 1–2 hours/day of monitoring, QA, and iteration
- •Best orchestrators are quant-minded and enjoy segmentation and experimentation
- •Most orgs must grow this role internally; external hiring market is immature
- •Managing multiple agents means managing segmentation, routing, and conflict resolution
- •Autonomous ‘master agents’ orchestrating other agents aren’t reliable yet
- 1:08:09 – 1:20:36
What’s changing vs. staying the same: support transformed, productivity expectations explode
Jason summarizes the near-term transformation: support is already largely AI-run, SDR/BDR roles shrink rapidly, and everyone is pressured toward far higher output. He argues ‘being a people person’ is no longer enough; product understanding and delivery of ROI fast are what matter.
- •Support is already 50–80% AI-handled across vendors—permanent shift
- •Cadence SDR and inbound qualifier roles are most vulnerable within ~12 months
- •Sales reps must become radically more productive using AI tools
- •‘People person’ is insufficient without product fluency and technical objection handling
- •Customers increasingly expect ROI during pilots—sometimes even before signing
- 1:20:36 – 1:42:10
Closing advice: more work (not less), the incognito test, and honest job impact
In the final stretch, Jason frames AI as creating more work through higher throughput—and more opportunity for those who lean in. He recommends the ‘incognito mode test’ to find broken customer experiences worth fixing with agents, and urges leaders to be honest about job changes—often driven by backfill decisions rather than mass firings.
- •AI increases output, creating a ‘keep up with the agents’ workload
- •Run the ‘incognito mode test’ with a fresh account to discover painful gaps
- •Fix the biggest pain by deploying an agent—fast path to meaningful improvement
- •Job impact is often via not backfilling roles rather than direct AI-driven layoffs
- •Lightning round includes tool picks (Reve for images) and standout SaaStr talks
