NederlandsFrom AI idea to proof of value. Not yet another sandbox PoC.
70% of AI proof-of-concepts never reach production. KODIFY flips it around: first sharpen the business case, then build. Five steps, one scorecard, a pilot that measurably delivers ROI — in days, not months.
Most AI projects don't fail on the model. They fail on the business case.
A sandbox PoC proves something can work. A Business Value Driven POC proves it should — with numbers that support an MT decision. It's the difference between a technical demo and an investment carried by risk, finance and operations alike.
KODIFY combines a proven model for AI business cases with the pragmatic, senior approach we're known for. We cut what doesn't work (PoC without an owner, a use case without a baseline, AI without governance) and keep what actually reaches production.
On this page we explain the approach in five steps. At the end of the page is a worked example case — not as a blueprint, but to make the approach concrete.
From strategy to production — via one scorecard.
A proven framework, refined in practice: senior advice, small teams, phased investment. Each step has one clear result — and a go/no-go before we proceed.
Connect AI to strategy
No list of loose experiments — we start with where the organisation wants to go. What strategic goals, which customers, which risks? That produces the filter for what AI should and shouldn't do.
A longlist that scores on value
In one or two workshops we map candidate use cases onto the scorecard: time savings, quality, revenue, risk reduction. High scorers proceed — the rest go in the backlog, not the roadmap.
Proof of Value, not Proof of Concept
This is the tipping point. We don't build a technical demo — we run a pilot with real data, real users, a baseline and a pre-agreed measurement bar. Succeeds — scale. Fails — stop, don't muddle through.
ROI, costs and risks on the table
The pilot delivers the numbers; we build the business case that finance, IT and risk can sign off on. Including total cost of ownership, change cost, sensitivity analysis and — crucially — what it costs to do nothing.
Embed, measure, adjust
A successful pilot is a beginning. We embed governance, human-in-the-loop, adoption and quality — and keep measuring quarter by quarter whether the promised value actually lands. No 'sign and disappear'.
A PoC proves it can work. A PoV proves it must.
We're pragmatic about PoCs: they're not bad — they're just misused when they end up in a sandbox halfway through. The business looks at money, quality and risk. That language needs to be in the PoC process from day one.
That's why we work as standard with a Proof of Value: the same technical rigour, but with production data, real users and a measurement bar that finance, IT and risk agree on upfront. What seemed a 'success' in a sandbox can turn out to be uneconomical on a PoV — and that's precisely valuable information.
One AI Use Case Canvas — where decisions rest.
For each candidate use case we score six dimensions. Below is a completed scorecard as it comes out of a typical value-discovery — and on which the step 4 business case is built.
Scored candidate — value-discovery output
Time saving per file
Throughput per file down, consistency up through uniform checks. Indicative annual value is substantiated with the baseline measurement from step 3.
Linked to a growth or risk goal
The use case solves a measured bottleneck (capacity, quality, throughput) and maps 1-to-1 onto a strategic goal — no loose experiment use case.
Available, governable
Sources are available and can be unlocked in a governed way. Sensitive data requires a tight flow and anonymisation — work, but not a showstopper.
Human-in-the-loop required
Employees remain the decision-maker. We design AI as assistant, not decision-maker. Training and escalation paths built in upfront.
Privacy, quality, audit trail
Privacy-by-design and a complete audit trail. Compliance frameworks are established in step 1 (governance), before the pilot starts.
Pilot in days, not months
Phased: first limited dataset, then scale. No bespoke software in PoV phase — existing best-of-breed tools preferred.
Roodhals — KYC in Financial Markets
An independent player in Financial Markets with a growth agenda and a KYC process that can no longer keep up with volume. KODIFY is running a PoV pilot in three workstreams, phased over 12 months.
Process analysis, tool selection, pilot with human-in-the-loop, evaluation and scale-up advice.
Monthly update on relevant AI tools in Financial Markets — with commentary on risk and impact.
Quality principles, privacy assurance, quarterly updates for partners and annual review.
Tool selection · pilot with limited dataset · governance frameworks · 1st market update.
PDCA · impact measurement on quality and efficiency · quarterly updates · follow-on domains.
Structural embedding · formally establish governance · adoption evaluation · roadmap year 2.
Got an AI use case in mind? Don't start building — start scoring.
30 minutes with Ruben Gelok, CEO of Kodify. We put your use case through the same scorecard, and tell you honestly whether there's a PoV pilot in it. No sales deck, no coffee chat — just the numbers.
