Guide
Most AI strategies are forty slides that nobody can act on. The ones that get funded fit on a few pages, name the use cases, the money and the risks, and tell the board what will be different in twelve months. Here is how to write one.
By Christopher Fagan, Managing Partner, PEF Consulting Partners
An AI strategy exists to answer three questions a board will ask: where will AI make us money or save it, what will it cost and what could go wrong. Everything else, the vision statements, the maturity models and the technology landscapes, is supporting material at best. If the document does not let a finance director approve a budget and a chief executive explain the plan in two minutes, it is not finished.
List the ten processes or decisions that cost the most, take the longest or create the most risk in your business. For each, ask whether the work is pattern recognition, language, prediction or judgement, because AI is good at the first three and poor at the last. Score each candidate on value, data readiness, risk and how hard it would be to run after the pilot. The strategy is the top three to five, in order, with a reason for each. Resist the temptation to include everything: a strategy with twenty use cases is a list, and lists do not get funded.
Every use case depends on data you may not have in a usable form, and on people who will have to work differently. Say so. A realistic AI strategy includes a data readiness view for each use case, the governance you will put in place before anything touches customers or staff, and the skills you will hire, buy or build. The organisations that struggle are not the ones that lacked ambition. They are the ones that assumed the data was clean and the staff would adapt.
A fundable strategy has a budget by phase, a first use case in production within six months, measured benefits by twelve and a decision point at each stage where the board can stop. It names an accountable executive and says what will be bought from vendors, what will be built and what will be run by whom afterwards. Across the AI programmes we have delivered, from a few hundred thousand pounds to tens of millions, the common feature of the ones that were funded quickly was a first phase small enough to approve without a committee and large enough to prove something.
Five to ten pages, written in plain English, with the use cases, the money, the risks, the governance and the ninety day plan. Review it quarterly, because the models, the vendors and the regulations will all have moved. A strategy that was right in January and unchanged in December was not being used. If you want a quick view of where you stand before you start, the free AI readiness check takes a few minutes and tells you which parts of the strategy will be easy and which will not.
Related reading: AI consultancy in London, AI readiness, governance and implementation consultancy, the free AI readiness check, a practical AI governance framework, AI consulting explained, where to start with AI when your business feels lost.
The business outcomes it serves, the three to five use cases in priority order, the data and skills each one needs, the governance and risk controls, the budget by phase, the owner and a ninety day plan. If it does not include money and names, it is a vision document rather than a strategy.
Four to six weeks for a mid sized organisation if the discovery is done properly: interviews, a look at the data, a review of what is already in use and a short options appraisal. Longer than that usually means the strategy is being written to avoid a decision.
You need both, in the right order. One or two well chosen pilots teach you more about your data and your people than any document, but without a strategy the pilots stay pilots. Write the short strategy, run the first use case inside it, and revise the strategy with what you learn.
A senior executive with authority over budgets and operations, supported by technology, data, legal and the owners of the processes being changed. If it sits only with IT it becomes a technology plan, and if it sits only with innovation it never reaches production.
If your AI strategy is stuck between a slide deck and a budget, a short confidential conversation will tell you what is missing.