Ask a founder how the founding AE interview went and the answer is usually about presence. They were impressive in the demo. They handled the pricing objection well. They asked good questions.

All true, and none of it predictive.

The founding seller at an AI company is hired into a job where the deck will be rewritten three times before their second quarter, the ideal customer profile is a hypothesis, and the buyer knows more about the technical problem than they do. Polish is table stakes. The skill that decides the outcome is something else entirely, and almost nobody interviews for it directly.

The Job Is Discovery, Not Delivery

A conventional account executive inherits a motion. Someone else has established who buys, why they buy, what they object to and what beats the competition. Execution is the job, and a demo-led interview tests execution fairly well.

A founding account executive inherits a founder’s intuition and a short list of design partners. Everything else is unresolved. Which segment converts, which use case lands first, which objection is fatal and which is noise, whether the technical evaluation is a two-week proof of concept or a two-month one.

The first ninety days should produce answers to those questions. Not a number.

Which means the interview needs to test how a candidate generates and pressure-tests a hypothesis, under conditions where nobody can tell them whether they are right.

Four Exercises That Actually Surface It

These replace the demo, or sit around it. Each takes under thirty minutes and each produces something comparable across candidates.

• The ICP teardown. Give the candidate three anonymised closed-won accounts and two closed-lost ones, with rough firmographics and the reason recorded. Ask them what pattern they see and which segment they would concentrate on for the next quarter. Strong candidates argue with the data and name what is missing from it. Weak candidates summarise it back.
• The reverse demo. Instead of asking the candidate to present your product, ask them to present a product they sold previously to you as a technical buyer, then interrogate it hard on cost, reliability and failure modes. You learn how deep their fluency runs when the script stops protecting them.
• The losing deal. Ask for a deal they lost that they still think about, and keep going past the first answer. The useful signal is whether the diagnosis is structural, such as the wrong buyer or the wrong stage, or whether the loss is attributed to procurement, timing and budget every time.
• The objection they cannot answer. Give them a genuine, live objection your product does not yet have a good response to. Say so. What you are testing is whether they sit with it honestly or reach for a deflection, because at seed stage they will hit it in week two without you there.

Between them these produce a much clearer picture than a polished demo does, and they are harder to rehearse.

The Technical Bar, Set Correctly

Founders selling AI and infrastructure products tend to swing between two errors on technical depth. Either they waive it, on the grounds that a good seller can learn anything, or they set it so high that only ex-engineers qualify.

The workable bar sits between the two. A founding seller needs to hold a first conversation with a platform lead without a specialist on the call. That means understanding what the product does, what it costs to run, how a buyer evaluates it and where it breaks.

It does not mean writing code, and it does not mean matching a staff engineer on architecture.

The practical test is simple. Have an engineer from your team spend twenty minutes with the candidate after a short product briefing, then ask the engineer one question: would you have let this person run that call without you?

That question has never produced an ambiguous answer.

Where the Strongest Candidates Come From

Founders tend to search inside the obvious pool, which is sellers with the right logos on the CV. It is a reasonable starting point and a poor stopping point.

Some of the strongest founding sellers in AI and infrastructure arrive from adjacent seats. Solutions engineers who have run the technical half of the evaluation and want the commercial half. Developer relations people who have spent three years earning credibility with exactly your buyer. Technical customer success leads who have seen what happens after the close, which makes them unusually good at qualifying out.

Each of these brings the harder half of the job already solved. Commercial mechanics can be taught inside a company with a competent founder. Technical credibility with an engineering buyer takes years and cannot be coached in a ramp plan.

Screening purely on quota history filters most of these people out before the first call.

The counterweight is real and worth stating plainly. Someone moving from solutions engineering into a founding seller seat has never owned a number, never run a commercial process end to end, and may find the first quarter of rejection harder than they expect. That risk is manageable, but only if it is named in the interview rather than discovered in month four.

So the question to put to an adjacent-seat candidate is not whether they can sell. It is what specifically drew them to the commercial side, and what they have already done about it without being asked to. People who have been quietly closing deals from a technical seat for two years will have an answer ready. People who are drifting towards the money will not.

What the Scorecard Should Say

Most founding AE scorecards are inherited from a later-stage company, which is why they measure the wrong things. A workable one for seed and Series A has four lines.

• Hypothesis quality. Can they form a defensible view from incomplete information and change it when the evidence moves?
• Technical credibility. Would your engineers let them run a first technical call alone?
• Build instinct. Is there evidence they have created something durable, such as a qualification framework, a competitive teardown or a repeatable discovery structure, rather than only having worked inside one?
• Tolerance for ambiguity. How do they behave when the answer genuinely is not known yet? The objection exercise usually answers this on its own.

Quota attainment still belongs on the sheet. It just belongs lower down it than most founders place it, because it measures performance inside a system that your company does not have yet.

The Broader Point

Interviews test what they are designed to test, and most founding GTM interviews were designed for a different company at a different stage.

The demo measures how well someone delivers a message that already works. The founding seller’s actual job is to find the message, prove it holds outside the founder’s mouth, and hand back a motion someone else can run.

Those are not the same skill. They are not even close.

Test for the one the role needs, and the shortlist changes considerably. So does the hire.

If you are about to open a founding GTM search and want a second view on the brief and the assessment before it goes to market, I would be happy to talk it through. You can see how we approach this on our page for hiring founding GTM and sales at AI startups.