We just moved 10+ years of data off Marketo into Salesforce Marketing Cloud.
I want to be clear that this wasn’t a decision we set out to make. A renewal came up, it became obvious they didn’t want us on the platform, and our hand got forced. But what happened in the weeks around it is the part every B2B vendor should be paying attention to, because the churn decision didn’t come from a procurement review. It came from an agent.
What happened
- Our agent started erroring out against their API. Roughly an hour a day of usable API, then it stalls. We couldn’t even run analytics against our own data.
- I asked the agent what we should do about it. Not rhetorically. It’s the same thing I’d ask a VP.
- It told us to leave Marketo, and gave us three more agent-friendly options. Reasoned, specific, hard to argue with.
- The support was the worst of any vendor we work with. Ten years as a customer and it was still the worst.
- They wanted another 12% after five straight years of increases. They were our single most expensive vendor.
- They never once offered to fix it. We would have signed at $20K with higher API limits. They had multiple chances to name $25K.
- We left. The migration took a week, cost about $14 in agent time, and I would never have voted for it without the agent.
Every one of those bullets is a decision a human used to make. Only two of them still were.
The API is now a churn surface
For twenty years, API quality was a sales objection at best. Something a solutions engineer handled on a technical evaluation call, then nobody thought about again until an integration broke.
That’s over.
Once your agents are connected to your data, the volume of queries changes by an order of magnitude. I used to ask our old finance person for a piece of analysis and wait a year for it. Sometimes I never got it. Now I have a question and I want the answer in 60 seconds, so I ask thirty questions a day instead of three a quarter. Everyone who has thrown a spreadsheet into Claude knows this feeling. You don’t ration your questions anymore.
An API budget built for nightly syncs is not an API budget built for an agent that’s actually working. Marketo gave us about an hour a day and then stalled out. That was survivable when a human was clicking through a UI. It’s fatal when an agent is the primary interface to your product.
And it doesn’t fail politely. The agent doesn’t file a support ticket. It reports that it’s blocked, over and over, until you ask it what to do about it.
Your agents will make the churn recommendation before your team does
An agent hits your limits. It surfaces the failure. A human, who is now spending most of their day working with that agent, asks the obvious follow-up question. And the agent, which has no relationship with your CSM, no memory of the deal you cut three years ago, and no political reason to protect the incumbent, gives a clean answer with alternatives attached.
That’s not a procurement process. There’s no RFP, no committee, no six-month evaluation. There’s a blocked agent and one question.
None of the old churn protections help. Switching costs are collapsing, and we moved ten years of data in a week. Relationship equity doesn’t apply to something that has no relationships. Familiarity is the last one standing, and I’ll come back to it, because it surprised me most.
What Marketo could have done for $25,000
If Marketo had come back and said, stay at $20K and we’ll raise your API limits, we would have signed. They had multiple chances to name $25K and take money we were prepared to hand them. Nobody ever did.
They may have needed to fix the API to make good on it. Maybe not even that. What they needed was somebody paying attention to the fact that their customer’s agent was hitting a wall, and the willingness to trade a price increase for a retention.
Instead they asked for 12% more on top of five years of increases, on a product our agent had already flagged as the constraint. A lot of B2B products we’ve used for five or six years now cost two or three times what they did. If you have no choice you might pay it. We had a choice, because the agent found one.
Two things I’d take from this if I were selling.
Support is still the cheapest retention lever in B2B, and the bar is still on the floor. Ten years in and Marketo was the worst vendor experience we had. Anybody can beat that. Some of the energy that used to go into post-sales has gone into forward-deployed engineers instead, and it shows in the accounts.
If your product delivers half its value once the agent is doing the work, you may have to let customers renew at half the price. Temporarily, maybe. If I’m getting the analytics value outside your platform and less inside it, charging pre-agentic prices for post-agentic value is how you lose the account to an agent’s recommendation instead of a human’s.
Mental contract length has collapsed to a year
Something related that I see across the agentic startups I’ve invested in.
They’re closing big deals fast. $50K, $100K and up, sometimes on a Saturday. And the founders are stressed, which sounds strange until you talk to their CROs and the veterans on their teams.
Those customers are making one-year commitments, and not the way everyone always has on paper. They’re doing it in their heads. The CRO signing your contract is telling themselves that the world is changing quickly enough that this might be the right agentic sales tool or marketing tool today, and they are not mentally committing to 2029 the way they would have five years ago.
So every one of those fast big deals carries renewal risk that isn’t showing up in anyone’s model yet. Your team knows it. That’s what the stress is.
The counterweight: you still can’t swap vendors constantly
I don’t want to oversell the switching-costs-are-dead story, because we lived the other side of it.
Yes, the migration was possible. Yes, the agent did the heavy lift in an hour for about $14. It still took a week of wall clock, because we kept hitting Marketo’s API ceiling on the way out. Amelia spent two solid weeks as our infrastructure person, and her verdict afterward was that she is not voting to spend the next two weeks the same way.
In the enterprise, a lift like this used to be five years of planning. Now it’s weeks of planning and a couple weeks of execution. Radically faster, but not free, and not minutes. You need planning, mapping, and a strategy before anybody executes.
So the practical limit is about one core vendor swap a year. Small bolt-ons that take five minutes to try, unlimited, and with more budget than in the past. Core systems, one. Even when it’s easier, doing it right is draining.
The other thing we got wrong is worth admitting. We dreaded this migration so much that we did nothing for a week. IWed used that UI for ten years. Amelia could build a campaign and send an email in 30 seconds. That fluency felt like a reason to stay.
The UI didn’t matter. We’re a week past the cutover and I wouldn’t go back. Whatever equity you think your interface has built up with your power users, an agent that can actually reach your competitor’s data will spend it faster than you expect.
Top 5 Takeaways
- Audit your own vendors on API quality now, before your agents do it for you. They will, and the recommendation to leave will arrive fully reasoned with alternatives attached.
- API limits are a retention surface, not a technical spec. An allowance built for nightly syncs breaks the moment an agent becomes the primary user of your product. That break is visible to your customer every single day.
- Trade price for retention while you still have someone to negotiate with. We’d have paid $20K to stay. Nobody asked. An agent’s churn recommendation doesn’t come with a save call.
- Best-in-class support is still the cheapest moat in B2B, and the bar is still low. Ten years as a customer and this was the worst vendor experience in our stack.
- Assume every customer is mentally on a one-year deal. Whatever the paper says. Price, support, and roadmap accordingly, because the agent doing the evaluating has no loyalty to you at all.
We went through all of this live on episode 11 of The Agents. If you want us to go deeper on any of it, tell me on LinkedIn or email jason@saastr.com. That’s my real email.
