the Business
No engineer has ever worked there.
US firms with $5–25M revenue (Census SUSB, 2022) — real payrolls, real operations, and almost none employ an engineer.
their entire technology relationship: an MSP — outsourced IT maintenance — at ~$36–63K/yr for a 30-person firm, patching and backups. Zero build capability on the premises.
Two surveys, people vs firms — but the direction is unmistakable: an implementation gap, not an awareness gap. The divide from class 02, standing in a warehouse. Closing it is the job.
Stated requirements lie politely.
“Often what the customer describes in scoping doesn’t match the data and system reality on the ground.” — Colin Jarvis, OpenAI, Head of Forward-Deployed Engineering
Build against observed pain: “you capture the tacit knowledge of how they work, not just the flattened ‘list of requirements’ model.” — Qureshi, eight years a Palantir FDE. That capture is what the weekly visits are for.
Most forward-deployed time is getting access to data, cleaning it, and integrating it. When you evaluate a problem, grade the pipeline, not just the demo — where does the data live, and will they let you touch it?
Ask about their life — not your idea.
Talk about their life, not your idea — an owner will politely bless any idea you pitch. Their daily reality can’t lie to you.
Three openers that always work: “Walk me through yesterday.” · “When did this last happen — what did it cost?” · “How do you handle it today?” Specifics about the past beat opinions about the future (“would you use…” invites a polite yes).
You’re there to listen. When something hurts: five whys down to the root, and shadow the work — watch the quote actually get made.
The stated problem was marketing. The real one is a three-day quote cycle standing on one spreadsheet. That’s the find. And notice who had the answer: Dana. Interview the people who do the work — not only the owner.
Some problems are AI-shaped.
Repetitive · rule-describable · data-touching. Internal utilities, glue between systems that don’t talk, document and data chores — the work someone does every Tuesday with 40 browser tabs open.
Payments, compliance, core customer data → established SaaS, full stop.If that IS their loudest pain, the deliverable becomes selecting and implementing the right SaaS — still a win, still scoped. (Class 15 makes this a rule with a checklist.)
Listen. Diagnose. Then scope one thing.
Connect.AI