A traditional AI strategy project takes 3–6 months, produces a 100+ page document and costs €100k–€300k. After all that, execution rarely starts because the strategy document doesn't tell you where to actually begin.
Building an AI strategy no longer needs 3 months, because industrial frameworks (12 industries pre-modelled with use-case libraries, semantic layer, governance) already exist. The work is not "invent something new" but "select and prioritise what fits our situation". Three other shifts: (1) EU AI Act is now known, so governance can be defined against regulation, (2) LLM tools have stabilised, so technology choices are less inflammatory, (3) real use cases have been seen in production — guesses can be replaced with data.
Week one measures AI maturity across eight axes: strategy, data, technology, skills, adoption, governance, culture and ROI measurement. Without a numeric T0 baseline, later progress cannot be proven.
Week two identifies 20–40 candidate use cases and narrows them to 6–10 prioritised targets. The prioritisation matrix uses four axes: business value, feasibility (data, tech, skills), governance risk and timing.
Week three defines the shared architecture on top of which selected use cases will be built: which data products (Bronze, Silver, Gold) are needed, what the semantic layer between use cases looks like, how agents are governed and how EU AI Act obligations will be documented. This week is the most critical: without shared architecture, every use case is built from scratch and unit economics never improve.
Week four compiles the roadmap, makes investment decisions and defines metrics. The roadmap covers 12 months and splits use cases into three waves: pilot (1–3 months), scale (3–9 months), production mainstream (9–12 months).