The interview is the moat

For years, organisations have treated interviews with subject-matter experts as an inconvenient step between commissioning a document and receiving it. Could the writer not simply use the existing material? There is a deck. There are some notes. Somebody made a SharePoint folder in 2019. Everything should be in there.

Everything is never in there.

The document contains the official process. The expert knows what happens when the process meets Tuesday morning. The slide says the service is flexible. The person delivering it can explain who is allowed to change what, how quickly, under which circumstances and what happened the last time a customer needed it. The case study says performance improved. The person involved remembers the argument, the failed first attempt and the tiny decision that actually made the difference.

That is the material.

AI makes existing language cheaper to rearrange. It does not make unspoken knowledge appear. If every competitor has access to the same tender, the same public sources and models of roughly similar ability, advantage moves upstream. The winning writer is not the person who can manufacture the smoothest paragraph from common inputs. It is the person who can obtain better inputs.

This is why interviewing is not a soft skill attached to writing. It is part of the intellectual work. A good interviewer hears the abstract answer and asks for the moment it became real. They notice the passive voice concealing a decision. They ask “compared with what?”, “who does that?”, “how do you know?” and, most profitably, “what usually goes wrong?”

The machine can transcribe the conversation, identify themes and help organise the evidence afterwards. Wonderful. Let it. The valuable bit happened when one person made another person think more clearly than they had before.

That is difficult to automate because it is not extraction. It is a relationship conducted in real time.