Why is Apex Replicant voice-first, and what does that mean for onboarding?

The problem this solves

Text-based AI clones ask you to upload everything you've ever written, then hope the model figures you out. That works for content volume. It doesn't work for tacit knowledge: the "why" behind the frameworks, the unwritten judgment calls, the things you'd only say out loud.

Voice-first onboarding gets at the tacit layer. You talk through your expertise the way you'd brief a new hire. The interview is structured (patent-protected evaluation across multiple dimensions), so nothing critical gets missed. What comes out is a protégé that sounds like you, thinks like you, and handles the judgment calls the way you would.

How it works

1. Application + intake. You apply. Our team reviews. If approved, you get a link to the interview.
2. The 60-Minute Method. A structured voice conversation covers your background, your methodology, the common client questions you answer, and the questions you wish more clients asked. The interview is the ingestion; content you upload later augments what we extracted here.
3. AI evaluation. The transcript is scored on multiple dimensions (patent-protected). If anything is thin, we flag it for you to fill in.
4. Auto-generation. Approval triggers our patent-protected generation pipeline. Your protégé's complete configuration is built in a single pass.
5. Auto-deploy. The generated protégé is pushed to our voice platform. Your test link is live.
6. You test it. Private sandbox. Ask the questions you get most often. Notice what's off. Submit feedback. Auto-regenerates.
7. You go live. Public, private (invite codes), or embedded. Your call.

How to turn it on

1. Apply at https://app.apexreplicant.com/signup (link from /for-experts)
2. Complete the 60-minute interview when your scheduled slot comes up
3. Review the post-interview evaluation email; fill any gaps we flagged
4. Approve the auto-generated protégé (or request a regeneration)
5. Test privately until you're comfortable
6. Go public, private, or embedded

FAQ

What if my expertise doesn't fit in 60 minutes? The interview extracts the tacit layer: your judgment patterns. It doesn't need to cover every fact. Facts get ingested into your knowledge base afterward (documents, audio, video, RSS).

Can I re-do the interview? Yes. Refinement feedback effectively re-interviews in writing. If you want a full second voice interview, ask us.

Does it capture my actual voice? Yes. Voice cloning from the interview. Your protégé speaks in your voice.

What if my protégé sounds like me but gets something wrong? Submit feedback in plain English. Our AI extracts instructions. The protégé regenerates and redeploys in the same request. Next session uses the improved version. (See /features/session-insights.)

How is this different from Steno's "Maya"? Maya launched in Steno 2.0 (February 2026). The structured interview protocol we use has been shipping since before our public launch; our protocol is patent-protected and on file. Steno's marketing describes upload-based training; ours describes interview-first extraction. The patent-claim difference matters if you care about long-term platform risk.