Amjad Masad on AI context windows

The kickoff half-day at SaaStr AI 2026 was the most concrete look we've put on stage at what running a company on AI agents actually looks like. Costs, bugs, drift, and the wins. Four sessions, four angles. 00:00 – AI Agents 101 with Jason Lemkin (SaaStr): building a digital clone on Delphi, the stair-step framework, buy vs build, guardrails, and how to price agents 1:10:00 – Building Business Agents at Scale with Jeanne DeWitt Grosser (COO, Vercel): the lead agent that took a team of 10 to 1, Deal One, the Playbook Platform, D0, Vertex, and a 32x ROI 1:41:00 – Build Your Own AI VP of Marketing, Live, with Amelia Lerutte (Chief AI Officer, SaaStr): how 10K went from a dashboard to a marketing co-pilot in 5 months 2:39:00 – Fireside with Amjad Masad (Founder & CEO, Replit): context engineering, the monorepo, self-improving agents, humans reporting to agents, and what deflationary economics mean for B2B What you'll learn: Why simple but trained beats complex What to buy, what to build, and why the line keeps moving How to set up two layers of agent autonomy with a human in the loop Why your data foundation and context are the whole game What breaks at production scale that never breaks in a demo SaaStr now runs close to 30 agents used nearly a million times, operating with a fraction of the headcount we had five years ago. This session shows the real stack behind it. More at SaaStr.ai Build your own with the spec + sample data: SaaStrAnnual.com/resources