Most advisors hand you a deck. Most builders have never run the org that has to live with what they shipped. I do both, which is the only reason the advice survives contact with your roadmap.
Most of my career has been spent somewhere the auditors got a vote — Barclays Global Investors, Absa, Skrill, Simply Business, TransferGo. Nine systems in production right now.
There is more underneath every claim on this page. The switch up there opens it.
Lightest first. Nothing here needs a procurement cycle to start.
A scored delivery-readiness position across five dimensions, the specific reasons your pilot hasn't reached production, a prioritised fix list, a build-or-buy call with the reasoning shown, a 90-day plan, and a one-page version your board can read.
Not included: implementation, hiring, vendor selection, or anything requiring me to be in your building. If you need those, say so on the call and I'll tell you who to speak to.
Monthly access fee, no timesheets, no day rate. One build seat at a time — if I'm full I'll tell you the month I'm free rather than take it and thin out.
Equity-inclusive structures are possible where the cash isn't, and I'll say plainly whether I think your equity is worth taking.
I'm based in Cape Town and I'm not relocating. SAST overlaps the UK and EU almost entirely and the US East Coast partially; if you need someone in a San Francisco room, I'm the wrong person and I'd rather say it here than in week three.
Paste a job description into the assessment and it will tell you where I fit and where I don't, with the gaps named.
Every claim carries a confidence rating and the limit of my own part in it. The ratings come from my evidence base, not from a marketing meeting — and I did not give myself five stars.
TransferGo revenue, 117% year on year, after the growth system I built as Product Director, Growth.
Localised landing pages across multiple languages as the substrate for A/B and multivariate testing, which moved acquisition off rising PPC costs and onto organic long-tail. A full onboarding rebuild for activation. Incentive structures for habituation — which compounds in remittance, because the need recurs.
And the part that mattered most: a GrowthHackers-style experimentation scorecard on Sean Ellis methodology, plus the data team that finally closed the loop on customer lifetime value — so cohorts could be tied back to unit economics. That is what made the $50m Series C raisable in 2021.
Verifiable by: Gareth Knight, then CGO.
Automated FAQ interactions served by Absa Chat Banking, on NLP architecture I commissioned.
A Camvas deck roadmapping intelligent summarisation, semantic search and automated transcription — before the ecosystem commoditised all three.
Front-end developer on the Barclays Global Investors team that built the first real iShares platform.
XML/XSLT and JSP on ATG Dynamo, back when decoupling data from presentation was still a live architectural argument. The payoff was that a new fund needed a new data feed, not a new page.
I've been making that same call ever since — most recently choosing pgvector
over a graph database on InfoTouch, and writing down the reasoning at the time rather
than after it worked. That through-line is the claim. The assets-under-management
numbers that platform later attracted are not mine and I won't borrow them.
My career evidence lives in a single controlled vault. Every proof note carries
attribution: and confidence: as structured fields, written when
the note was created — not when someone asked an awkward question. The stars are those
fields, rendered.
That's why one of them is two stars. A rating system where everything scores well isn't a rating system, it's a brochure.
Not a compliance audit. The question is whether your AI work reaches production and survives contact with the organisation — team, workflow, data, who owns the decision, and how it's governed. A conversation, not a form; it follows up on what you actually say. Score, read and plan are free. Only the full roadmap asks for an email.
"Your pilot works and it's been three months. What has to be true for someone other than the team that built it to depend on it — and who decides that?"
Five dimensions: team, workflow integration, data readiness, decision ownership and governance. Four of them are delivery questions and I score them on my own judgement, which is what you're here for.
The governance dimension is the one with an external anchor: each question maps to a named subcategory of the NIST AI Risk Management Framework 1.0 — four functions, nineteen categories, seventy-two subcategories. Free, public, and you can read the exact line I scored you against.
The 1–5 levels are mine, not NIST's. NIST deliberately does not define maturity
levels — it is explicitly "outcome-focused and non-prescriptive" and says its actions
"do not constitute a checklist". Anyone selling you a "Level 3 against NIST" invented
that ladder too. The difference is that mine is published:
0 / 1 / 2 per subcategory, mean × 2.5.
I don't reproduce ISO/IEC 42001 — it's paywalled at about CHF 225, which is exactly why I won't pretend you can verify a score against it. If what you actually need is a regulator-defensible governance audit rather than a delivery read, say so and I'll point you at the people who do that properly. It isn't this.
I will not tell you how you compare to your peers. No public dataset supports that percentile and I don't have a respondent population yet. When I do, I'll publish the n. Every "+2.2 above average" you've seen on an assessment like this was made up.