SaaStr ditches Marketo after AI agents hit API limits, migrates in one week
## The churn signal came from an agent, not a human SaaStr moved 10+ years of marketing data off Marketo after their AI agents kept hitting API rate limits. Not a strategic review. Not a procurement committee. An agent errored out, they asked it what to do, and it told them to leave. Jason Lemkin runs SaaStr (the SaaS media and events business) with 3 humans and 20+ AI agents. The agents were querying Marketo's API constantly for analytics and workflow automation. Marketo's limits allowed roughly an hour of usable API time per day, then stalled. When you are running agent-driven ops, that is not a minor inconvenience. That is a blocker. Lemkin asked the agent what they should do about it. The agent gave three alternative platforms with better API limits. All more agent-friendly. Hard to argue with. ## What Marketo could have done Marketo was their most expensive vendor. Five straight years of price increases. Another 12% on the renewal. Support was the worst of any vendor they work with, according to Lemkin. He says they would have stayed at $20k with higher API limits. Marketo had multiple chances to name $25k and keep the account. Nobody did. Instead, they pushed for another increase on a product the agent had already flagged as the constraint. The migration took one week. Cost $14 in agent time. Lemkin says he would not have voted for it without the agent recommendation. ## Why this matters for sales tools API quality used to be a solutions engineer problem during evaluation, then forgotten. Now it is a churn surface. When agents are your primary interface to a product, query volume goes up by an order of magnitude. A budget built for nightly syncs does not work when an agent is asking 30 questions a day instead of 3 per quarter. The agent has no relationship with your CSM. No memory of the deal you cut three years ago. No political reason to protect the incumbent. It surfaces the failure, gets asked for alternatives, and gives a clean answer. Switching costs are collapsing when agents do the migration work. Relationship equity does not apply to something with no relationships. Support quality and API reliability are now retention levers, not nice-to-haves. Lemkin's take: if your product delivers half its value once the agent is doing the work, you may need to let customers renew at half the price. Charging pre-agentic prices for post-agentic value is how you lose accounts to an agent's recommendation instead of a human's. ## The comp angle SaaStr previously had a traditional sales team. Lemkin has publicly said he replaced most of them with AI agents and stopped hiring humans for sales roles. That is 20+ agents doing work that used to require headcount with OTEs. For sales professionals: this is not a vendor case study. This is what happens when your buyer's ops team runs on agents instead of people, and those agents start making vendor decisions based on API performance, not relationship history.