AI Readiness in practice: where to actually start

Almost every company we talk to wants "to do something with AI." Far fewer can answer the question that decides whether it works: is the business actually ready for it? The gap between a flashy demo and a tool people rely on every day is rarely the model. It is the data, the process and the people around it.

Over dozens of projects we have settled on a five-step readiness assessment. It takes a couple of weeks, costs a fraction of a full build, and tells you exactly where the value is, and where the landmines are, before you commit a budget.

1. Find the use case worth the effort

Not every process is worth automating, and the most visible one is rarely the most valuable. We look for tasks that are frequent, rule-heavy and time-consuming, where a small percentage improvement compounds into real hours saved.

• How often does this happen, and how many people does it touch?
• What does an hour saved here actually cost today?
• Is there a clear, measurable outcome we can hold the pilot to?

2. Audit the data

AI is only as good as what it can see. Before anything else we map where your data lives, how clean it is, and who is allowed to touch it. This is where most "AI projects" quietly die, not because the model cannot reason, but because the inputs are scattered across five systems and a shared drive.

"If your knowledge only exists in people’s heads, the first AI deliverable is not a model, it is writing it down."

3. Map it to the process

An assistant that sits in a separate tab nobody opens delivers nothing. The assessment includes a hard look at where AI fits into the existing workflow, what triggers it, who reviews its output, and what happens when it is unsure. The best implementations are almost invisible: they live inside the systems people already use.

4. Ship a focused pilot

We never start with a platform. We start with one real use case, a clear KPI and a two-to-four week timebox. The pilot answers a single question: does this measurably beat the current way of doing it? If yes, we scale. If no, we have spent weeks instead of months learning that, and we tell you straight.

5. Plan for production from day one

A demo runs once. Production runs at 5 a.m. on a holiday. From the first sprint we think about monitoring, guardrails, cost, and the people who will own the system after we hand it over. AI in production is software, and we treat it with the same discipline as everything else we build.

The honest version

Sometimes the assessment concludes that the highest-value move is not AI at all, it is digitalizing a process first, or fixing the data foundation. We will tell you that too. AI as an accelerator, not decoration, is one of our core principles, and readiness is how we keep that promise.