We just shipped the biggest update yet to our AI Agent Playbook for GTM. It is the October 2026 edition: 53 pages, 27 parts, every number, every mistake, and every framework from running SaaStr on 3 humans and 21+ AI agents in production.
If you only know the June version, this is a near-total refresh. We folded in everything from SaaStr AI Annual 2026, the latest episodes of The Agents, the ICONIQ 2026 GTM benchmarks, and a round of deep dives with operators actually doing this at scale. No theory. Receipts.
What we accomplished with 21 agents:
- $4.8M in pipeline sourced via agents, and $2.4M closed-won first-touch from agents
- Deal volume more than doubled, win rate nearly doubled, and none of it cannibalized other revenue. It was all additive
- 614 meetings booked by a single inbound agent at one event, at a ~$85K average deal
- Two of our AI VPs, 10K and QBee, cost us $257 last month, combined
- Revenue went from -19% to +47% year over year with the same three people
What’s new in this edition:
- The agent evolution pattern. Almost none of our agents started as agents. They started as a dashboard, a tracker, or a website, and became agents because we showed up to work with them every day. This changes how you should start.
- Headless CRM. The one-hour win most teams are skipping, plus why great APIs can make dated software competitive again.
- The A/B/C/D lead rule. Stop pointing AI SDRs at your A-leads. The scored, ignored B-leads are where the found money is. For us that was an extra $500K off one event.
- Field lessons from running the stack. Why 30 minutes a day became 8 hours a day. Leaving Marketo after 10 years. The day an agent picked a vendor we had never heard of and a paid competitor was never even in the deal. Why high usage is not loyalty, and why stealth churn now cuts both ways.
- The market data. We analyzed 22,000 director-and-up execs from our events: AI-native titles are up 4x, and half changed jobs in the last 16 months.
- Six new case studies. Owner.com, Klaviyo, ElevenLabs, Harvey, Rubrik, and Glean, with the real numbers behind each.
What’s inside
We grouped the 27 parts into four sections so you can jump to what you need:
- Section 1 — Foundations. The numbers, the key insight (pretty good at scale beats perfect), the #1 error everyone makes, and how to copy your best human.
- Section 2 — Build & Operate. The full 21+ agent stack, the training reality nobody talks about, the 90-day deployment framework, the mistakes that kill AI SDRs, the vendor reality, the economics, and how we built our own AI VP of Marketing.
- Section 3 — Proof From the Field. Every case study in one place, the ICONIQ 2026 benchmarks, and the hardest lessons from The Agents podcast.
- Section 4 — Field Lessons & What’s Next. The newer deployment tactics and the operator Q&A on planning, comp, and hiring.
We’ve moving to an agent-first world
A year ago, running a couple of agents took us about 30 minutes a day. This week it was eight hours a day, and we are busier than we were with a 20-person team. The reason is the whole story of this edition: when an agent only does a task, you check the task. When an agent makes decisions, you need an opinion on every decision it makes, and it does not tell you about all of them. A year ago they failed loudly. Now they hand you something that looks finished. That shift is where all the leverage, and all the new risk, lives.
Where to download
It is free. Grab the latest edition at saastr.ai/agents, where we also keep the full agent stack, the decks, and every session updated as we go.
