Most companies treat a first sales call as exactly that — a sales call. Ours is closer to a structured audit, and we tell prospective clients that upfront, because it changes what they should expect to walk away with, and it shapes everything we build after.
What we're actually trying to learn
The goal of the discovery call isn't to pitch Wrennon. It's to figure out, honestly, whether an AI agent is the right fit for the specific process a team wants to automate — and to be direct if it isn't. We ask about ticket volume, but more importantly we ask about ticket variety: how many genuinely distinct question types show up, and how much of that variety follows a pattern versus requiring real human judgment.
We also ask questions most sales calls skip entirely — what's the worst version of this going wrong, what would erode customer trust fastest, and what's never allowed to happen regardless of how rare it is. Those answers shape the guardrails and escalation rules built into the agent long before a single prompt gets written.
Why we sometimes say it's not a fit
Not every process should be automated, and we'd rather say so on the first call than take on a project that was never going to succeed. A support queue that's mostly genuinely novel, judgment-heavy questions with little repeatable pattern isn't a good early candidate for an agent, no matter how much a team wants to reduce ticket volume. We've turned away work on this basis, and we think that honesty is part of what makes the projects we do take on succeed.
How this shapes the build that follows
Everything from the process mapping sessions to the evaluation test set traces back to what surfaces on this first call. A team that tells us their biggest pain point is post-purchase shipping questions gets a very different initial build focus than a team whose real bottleneck is pre-purchase product comparison questions — even if both teams initially described their need as "we want a chatbot."
Treating this call as an audit rather than a formality is slower than jumping straight into a proposal. But it means the scope we agree on is grounded in what a team's process actually needs, not in a generic package we'd sell to anyone who asked.
What we send after the call
Most sales calls end with a follow-up email and a proposal. Ours ends with a short written summary of what we learned — the process areas discussed, the edge cases already mentioned, and an honest note on which parts look like a strong fit versus which parts need more discovery before we'd commit to a scope. Clients tell us this is unusual, and also that it's the reason they trusted the eventual build to reflect what they'd actually described, rather than a generic pitch reworded with their company name inserted.
It also means that if a client decides not to move forward, they still walk away with a genuinely useful artifact — a clearer picture of where their own process has undocumented exceptions, independent of whether they ever build an agent for it.