What is the 60-Minute Method?

Drew Harris · CEO and Chief Product and Technology Officer · 2026-04-08 · 7 min read

methodologyonboardingzero-hallucinationpii

Why the interview length matters

Most AI clone platforms ask for content. Upload your book, your blog archive, your YouTube transcripts, your course material. The model does its best to reconstruct you from the exhaust.

That approach has a ceiling, and the ceiling is tacit knowledge: the "why" behind your frameworks, the unwritten judgment calls, the things you'd only say out loud. You don't write these down because you don't think to. You say them when a client asks the right question.

A structured voice interview is the fastest way to surface tacit knowledge. The interviewer asks the right questions on purpose. You answer the way you'd answer a sharp new hire. The protocol makes sure nothing critical gets missed.

Sixty minutes is the length where the interview stops being polite small-talk and starts being useful, and hasn't yet hit the point where you're repeating yourself. It's not arbitrary. We tried thirty; it was thin. We tried ninety; experts tired out and the last half-hour added nothing the previous hour hadn't covered better.

What the interview actually covers

The interview is conducted voice-first by an AI interviewer. Your responses are transcribed in real time. The conversation follows a protocol but isn't a script; the interviewer adapts based on what you say.

The protocol covers four sections:

Section 1: Your background and your methodology

Where you trained, who shaped your thinking, what you teach. The goal is to place your expertise in a lineage, not for marketing, but because a lineage-aware protégé handles "where does that come from?" questions differently than a rootless one.

Section 2: The frameworks you actually use with clients

The mental models, the checklists, the sequences, the decision trees. Experts often have five or six of these. Most have never written them all down in one place. The interview is the first time they do.

Section 3: The questions you get most (and the ones you wish you got more)

The common questions become the protégé's bread-and-butter answers. The wished-for questions are often more useful; they're where your depth is, and a client who doesn't know to ask them is the client you most want to impress.

Section 4: The judgment calls

Edge cases. When do you say no? When do you refer out? What separates advice you give to a retainer client from advice you'd give to a stranger on LinkedIn? The judgment calls are the hardest to write down, the most valuable to capture, and the thing least likely to appear in a bot built from your blog posts.

The post-interview evaluation

After the interview, the transcript runs through an AI evaluation. The evaluation is patent-protected and proprietary; what you see, as an expert, is a pass/fail with targeted feedback on any areas that came through thin. Common feedback themes: clarify a framework that wasn't fully surfaced, add color on the edge cases your clients don't usually ask about, or schedule a brief follow-up to fill a specific gap.

Thin evaluation scores don't block the protégé, they trigger a short fill-in-the-gaps pass before generation. The evaluation is a safety check, not a grade, and the purpose is to prevent a protégé from shipping with silent holes.

The evaluation is how we catch the silent failure mode: a perfectly pleasant interview that didn't actually surface the thing that makes you distinctive.

From interview to deployed protégé

The interview is the single longest step, not the total timeline. Here's what the rest looks like:

Interview (~60 min): voice conversation, transcribed.
AI evaluation (~minutes): evaluation scoring, gap flagging.
Admin review (same day): our team confirms the transcript is deployable.
Auto-generation (~minutes): our patent-protected generation pipeline produces the protégé's complete configuration in a single pass.
Auto-deploy: the protégé is pushed to our voice platform; you get a private test link.
You test: private sandbox; submit feedback the way you'd edit a junior's work.
Feedback loop (our refinement pipeline): your feedback extracts into structured instructions, regenerates, redeploys.
Go live: public, private (invite codes), or embedded on your site.

Fastest end-to-end deployment on record: James Buff, under 24 hours. That's interview → test → feedback cycle → live client sessions. Most experts take longer because most experts want to test more carefully, and we encourage it. But the platform is not the bottleneck.

Why sixty minutes, not fifteen and not four hours

Shorter interviews (15–30 min) produce thin transcripts. The evaluator flags everything. The protégé ships with known gaps, which either fill in through the feedback loop (fine but slow) or surface as "I don't know" during sessions more than they should.
Longer interviews (90+ min) don't add proportional value. Experts fatigue. The last half-hour restates the middle hour. Transcript quality drops.
One 60-minute session outperforms two 30-minute sessions in our testing; the flow of thought matters, and splitting the interview breaks it.

The method isn't the only ingestion event. You keep adding to the knowledge base after launch (documents, transcripts, RSS feeds, session summaries). The interview is what seeds the voice, the judgment, and the frameworks. Everything else augments.

What happens after the interview

The deliverable isn't a chatbot; it's an expert protégé with:

A zero-hallucination architecture enforcing that it only speaks from content you've provided
A configurable opening greeting in your voice
A structured-output schema for intake or assessment, if your use case needs it
Automatic PII protection on every ingest
Session insights after every session, where Claude extracts action items, unanswered questions, and refinement suggestions
A one-request feedback loop for ongoing tuning

The interview is the protocol. The rest of the platform is the infrastructure that makes the interview's output durable over time.

FAQ

Is the 60-Minute Method the same as the full deployment timeline? No. The 60-minute figure is the interview length. End-to-end deployment (interview → live sessions) is as fast as 24 hours (James Buff) and typically a few days for experts who test deliberately. A multi-week governance period is optional for experts who want maximum refinement before going public; it's not a required phase.

How is this different from Delphi's Interview Mode or Steno's Maya? Delphi shipped Interview Mode in November 2025; Steno launched Maya in February 2026 as part of the Steno 2.0 rebrand. Both are structured interview flows. Our protocol is patent-protected and pre-dates both launches. Competitor protocols aren't publicly disclosed at the same level of architectural detail.

Can I do the interview myself, or does someone guide it? The interviewer is an AI voice agent that follows the protocol adaptively. There's no human interviewer in the standard flow; the AI asks better follow-ups more consistently than a human would, because it doesn't get tired and doesn't skip dimensions.

What if my interview transcript scores low on evaluation? You get specific feedback on which dimensions were thin. You can either schedule a short follow-up interview to fill the gaps, or proceed knowing those will surface as "I don't know" responses until addressed. Most experts choose the follow-up.

Do I need to prepare for the interview? A light prep helps: think about the frameworks you actually use, the questions you get most, and the judgment calls that distinguish your work. The interviewer will guide the rest. Over-preparing tends to produce a lecture rather than a conversation; we get better transcripts from experts who come ready to talk, not ready to present.

Can the interview be done by text instead of voice? No. Voice is intentional. Spoken responses surface tacit knowledge that written responses lose. Experts explain things differently when they're talking.