Independent AI consultancy
The AI that's worth doing
Strategy. Development. Leadership.
Most AI projects fail quietly, somewhere between the pitch deck and production. We find the work that's genuinely worth doing, build it properly, and measure whether it moved anything.
yes, that is our logo moonlighting
§ 01 What we do
Three ways we plug in
Some clients need a map, some need builders, some need someone senior in the room every week. Most need two of the three, in an order that only becomes obvious once we start.
01 / STRATEGY
AI Strategy & Consultancy
Fig. 1.1 — Locating the minimum
Most AI roadmaps are wish lists sorted by excitement. We map your real constraints, find where the payoff is steepest, and tell you plainly which ideas to drop.
Read more02 / DEVELOPMENT
Custom AI Development
Fig. 1.2 — Two paths, recombined
Systems that survive contact with real users, real data and real load. Evaluated against a baseline before they ship, and monitored after, because that is when models quietly rot.
Read more03 / LEADERSHIP
Fractional AI Leadership
Fig. 1.3 — A stable configuration
Senior technical direction without a senior headcount. We set direction, mentor the people you already have, and keep the roadmap honest when the hype cycle turns again.
Read moreIn practice
The interviews stayed human. Two weeks of heavy analysis now takes twenty minutes.
02 / Development Perlon AIA working first version in 12 days. Their engineers took it from there.
03 / Leadership VericodeFractional AI lead since 2025. The AI on dott.ai was designed and built with their team.
§ 02 How we work
We run it like an experiment
Consulting borrowed most of its habits from sales and almost none from science. We prefer the other order: state what you expect, build the smallest thing that tests it, then check honestly whether it held.
-
01Observe
Look first
We sit with your data, your team and the workflow as it actually runs. A surprising amount of what everyone "knows" turns out never to have been measured.
-
02Hypothesise
Commit in advance
We write down what we expect, and the number that would tell us we were wrong. Agreed before the build starts, so nobody moves the goalposts later.
-
03Build
Smallest useful thing
The least machinery that can genuinely test the hypothesis, put in front of real users. Not a prototype that only works on the demo data.
-
04Measure
Report either way
Against the baseline we agreed on. If it worked, you get the number. If it didn't, you get that number too, plus what we learned and what we'd try next.
null results are still results
§ 03 A note from a client
"Prosop.ai's strategic and scientific consulting for AI integration has made all the difference for the Vericode and dott.ai products."
§ 04 Trusted by
Teams we've worked with
§ 05 Contact
Let's find out whether you have a problem worth solving
A first conversation costs nothing and is genuinely useful either way. Sometimes it ends with a proposal. Sometimes it ends with us telling you that you don't need AI for this, which is also a good outcome.
Worth putting in the first email
- The decision or process you wish were faster, cheaper or less painful
- What you've already tried, including the things that didn't work
- Roughly what data exists, and who owns it
- How you'd know, six months from now, that this was worth it
no NDA needed to say hello