n.V6-4.05 | THE ASSISTANT AS MENTOR
01| The economic case for agents is about cost. The human case is about capability, and it is the one I care about more. Every Ikosi who works with an assistant should come away more capable than they would have been alone. Not flooded with output. Not patronised. Not confidently misled.
02| Concretely, the assistant helps research and analyse a problem in the barrio or on the platform. It offers the right template for the work at hand — a project plan, a report, a proposal, a message to a buyer — instead of leaving a person to invent a format under pressure. It walks them through platform processes step by step. It surfaces what other Maveriqs did in similar situations, including what failed. It catches the errors and the gaps before the work goes out. It translates — between languages, between technical and plain speech, between platform vocabulary and how people actually talk on the street. It tracks progress and reminds people of the next step, because momentum dies in the gaps. And it knows when to stop and connect the person to a human being.
03| All of that depends on memory. An assistant that forgets you between sessions is a search box with manners. A mentor relationship requires that it remembers your history, your projects, the advice it already gave you, and what you did with it. Whether memory can be made reliable enough, for enough people, over years, is a load-bearing assumption and it is not proven. If it fails, we still have a useful tool and we have lost the mentor.
04| There is one behaviour I want built in from the start, and it runs against the grain of how these systems are usually tuned. When a question has two legitimate answers depending on context the assistant does not have, it must say so and ask, rather than pick the more likely one and sound certain. A Maveriq asking whether to run a project in the rainy season is asking a question where the answer depends on things the machine does not know. The honest response is a question back.
05| What we want: an assistant that a person can rely on the way they would rely on a good colleague — including relying on it to say "I don't know" and "that's outside what I do".
06| What we sacrifice: speed and polish. An assistant that asks before answering is slower than one that answers immediately, and in a demo it looks worse. We take the slower one. Certainty is cheap to fake and expensive to lose.
07| The real risk is not technical. It is rejection. If Maveriqs do not use their assistants — because it feels like surveillance, because the first answer was wrong, because it talks down to them, because it was introduced badly — then everything in this chapter is a slide in a presentation. Adoption is the test, and adoption is won on the ground by people, not by the feature list.
