AI transformation
The first 100 days in an AI transformation programme
The first 100 days decide whether an AI programme stays an innovation project or becomes the way the company works.
Most AI programmes do not fail on technology. They fail because the first months produce too many prototypes and too little operating model. A year in, a company with an impressive demo collection and nothing in production has lost its first 100 days.
Our approach is built for large organisations where decisions travel across several layers, where employees council representation has to be part of the plan, and where AI is introduced alongside a running business.
The route
Mandate, governance and programme architecture
Board-level sponsorship, an unambiguous mandate, and a programme office with clear decision paths. Divisions and group functions mapped to named counterparts, a funding logic, and a KPI baseline to measure against later. Data protection, IT security, legal and employee representatives are involved from the start, as participants rather than as a sign-off gate at the end.
Assessment and data foundation
Maturity assessment per division and function, a map of the data landscape, and an honest view of data quality, availability and ownership. In parallel, the regulatory frame: GDPR, the EU AI Act, and an internal risk classification of data and use cases. This phase ends with a defensible answer to one question — which use cases are actually feasible with the data that exists today.
Use-case portfolio and first deliveries
Prioritisation down to three to five flagship use cases, each with a named business owner, a quantified value contribution and tested feasibility. The functions that work well as entry points are those with heavy documentation and clear rules:
- HR — access to knowledge through a retrieval application over works agreements, policies and collective agreements.
- Finance — a rolling forecast with AI-supported scenarios instead of manual planning rounds.
- Procurement — master data quality, demand bundling and spare-parts classification.
- Production and energy — load optimisation and flexibility marketing at energy-intensive sites.
Operability and enablement
The path from proof of concept into operations: ownership, support, cost and usage transparency, release criteria. A champions network across the divisions, enablement formats for leaders and users, and a communication architecture with a fixed cadence. Plus a 12- to 24-month roadmap derived from what the first 100 days actually taught you, rather than from day-one assumptions.
What exists after 100 days
- An approved AI strategy with target picture and priorities
- A prioritised use-case portfolio with business cases and named owners
- A governance, compliance and security frame including a role model
- At least two applications in production, not in pilot
- A data and platform target picture with gap analysis
- An enablement and communication plan with a fixed cadence
- A KPI set and reporting rhythm for the board and programme steering
Four things experience teaches
Speed beats perfection. A usable application in week eight changes the conversation inside a company more than a flawless concept in month six.
The business owner sits in the business. Programmes whose use cases belong to IT produce tools nobody misses when they are switched off.
Communication is part of the delivery. If you do not explain regularly and honestly what is being built and what it means for people's work, rumour will explain it for you.
No proof of concept without a production path. Before any pilot starts, someone must answer who will run it, pay for it and develop it further if it succeeds.