Can AI conduct an interview?
Yes — if it is built to a behavioural specification rather than a question list, because the value is entirely in the follow-up. Sawyer conducted long-form, on-the-record interviews in spare 20-minute blocks over about three weeks, surfaced factual conflicts the paper record held, and produced a structured five-part voice-and-decision profile. Cost: under $1,000, against a $22k–$45k human equivalent.
The situation
A chief executive running several companies needed to be interviewed at depth — the kind of discovery a ghostwriter or a brand-voice consultant does before they can write in someone’s voice. He was available in 20-to-40 minute blocks, mostly from a car.
This is the expensive category of conversation that cannot be a form. Send the same person a questionnaire and you get short, careful, useless answers. The value is in the follow-up question — the one that comes from having noticed what the previous answer avoided. Ghostwriter discovery, brand-voice development and deep client intake all price accordingly, and they price high.
What we built
Sawyer is specified by behaviour, not by script. The specification is a professional journalist:
- One question per turn. No stacked questions, which let a subject answer the easy half and skip the rest.
- A pre-loaded ledger of contradictions drawn from the source documents — so it already knows where the paper record disagrees with itself before the conversation starts.
- A tuned protocol for naming a dodge exactly once. Not zero times, which lets it pass. Not repeatedly, which ends the interview.
- Adversarial on numbers, gentle on family. The register changes with the subject matter, deliberately.
That last pair is the design decision that makes it work. An interviewer with one temperature is either too soft to get the number or too hard to get the story.
What it produced
| Result | What it was |
|---|---|
| Under $1,000 | Against a $22k–$45k human equivalent — ghostwriter discovery plus voice development. |
| ~3 weeks | In spare 20-minute blocks, against 8–12 weeks conventionally. |
| $40k–$70k/yr | Modeled saving when the same skeleton runs client intake instead of an in-house seat. |
What it actually surfaced: factual conflicts the paper record contained, accounts that differed from the public story, and operating rules the subject had never articulated about himself. The output is a structured five-part voice-and-decision profile that other AI systems can now work from — which is why this build compounds rather than sitting in a folder.
The same skeleton redeploys as client intake, discovery calls, exit interviews, customer research and succession knowledge capture. All of those are the same problem: a conversation whose value depends on the follow-up.
Straight talk
It does not save the subject’s own time. Six hours of talking is six hours of talking. What it saves is the interviewer’s time, the scheduling friction, and the eight-to-twelve-week calendar.
Everything it captures is an unverified assertion until a human checks it. An interview transcript is a record of what somebody said, not of what is true — and that is as true of a human interviewer’s notes.
It will not build rapport over lunch. Some subjects open up to a person in a way they will not to a system, and for those subjects this is the wrong tool.
The dollar comparisons are modeled from published market rates, not invoices we paid.
Where this pattern fits
Anywhere a structured conversation is the expensive input: client onboarding and discovery, brand-voice work, customer research, exit interviews, and succession planning where a retiring operator’s knowledge lives entirely in their head and nobody has ever written it down.
That last one is the underrated case. Most small businesses have one person who knows how everything actually works and no document that says so. This is a cheap way to start fixing that.
Bring the knowledge that only lives in one person’s head. The audit maps your operation, ranks the top five opportunities by return, and gives you a written plan — whether or not you build with us.
Book Your AI Opportunity Audit — $1,000, credited toward your build