Many organisations have tried artificial intelligence in some form. A team experiments with a chatbot, a department buys a tool that promises automation, and staff start using generative AI on their own. Yet a year later, leaders often struggle to say what has actually improved. The problem is rarely the technology. It is the absence of a clear strategy that connects AI to the organisation's goals, data, people and risks.
An AI strategy does not need to be a thick document written by consultants. It is a set of decisions about where AI will create value, what must be in place for it to work, and how the organisation will manage it responsibly. The steps below apply to companies and government bodies alike.
1. Start with problems, not with technology
The most common mistake is to begin with a tool and look for somewhere to use it. A better starting point is the organisation's own priorities. What are the three to five outcomes leadership cares about most this year? Faster service to customers or citizens, lower operating costs, better decisions from existing data, fewer errors in routine work?
For each priority, ask where people spend time on repetitive tasks, where decisions are slow because information is scattered, and where quality varies from one employee to another. These are the places where AI is most likely to help. Write them down as business problems in plain language, for example "it takes too long to answer routine enquiries" rather than "we need a chatbot".
2. Assess your readiness honestly
Before committing budget, take an honest look at four areas.
- Data: Is the data you need available, reasonably complete and stored in a way that can be accessed safely? Many AI projects stall because data sits in spreadsheets, paper files or systems that do not talk to each other.
- People and skills: Do managers understand what AI can and cannot do? Is there anyone who can evaluate vendors, manage a pilot and measure results?
- Technology: Can your current systems connect to new tools, and does your IT team have capacity to support them?
- Culture and leadership: Are senior leaders willing to sponsor change, accept some experimentation and redesign processes rather than simply automate old ones?
A simple readiness score of high, medium or low for each area is enough. The aim is to know where the gaps are, not to produce a perfect assessment.
3. Choose a small number of high-value use cases
With the problem list and readiness picture in hand, select two or three use cases to start with. Good first use cases share some features: the benefit is clear and measurable, the data already exists, the risk is manageable if something goes wrong, and the people who will use the solution are willing to take part.
Typical examples include drafting and summarising documents, classifying and routing incoming requests, searching internal policies and procedures, forecasting demand or workload, and detecting anomalies in financial or operational data. Each organisation will have its own priorities, and it is better to do two use cases well than ten poorly.
For each chosen use case, define what success looks like before you start: time saved, errors reduced, response times improved or another measure that leadership already recognises.
4. Put governance and risk management in place early
AI introduces risks that need deliberate management: inaccurate outputs, bias, leaks of confidential information, and unclear accountability when a decision goes wrong. Governance does not have to slow things down, but it must exist from the first pilot.
At a minimum, agree on a short AI use policy for staff, decide which data may and may not be used with external tools, require human review for decisions that affect people's rights or money, and name an owner for each AI system. Government bodies should also consider transparency towards citizens and any national rules on data protection and digital government.
Good data governance is the foundation. Clear data ownership, quality standards and access rules make every later AI project faster and safer.
5. Build skills across the organisation
An AI strategy succeeds or fails through people. Senior leaders need enough understanding to set direction and ask the right questions. Managers need to know how to redesign work and lead teams through change. Specialists need practical skills in data, prompt design, evaluation and vendor management. All staff benefit from basic awareness of safe and effective use.
Plan learning at each of these levels and link it directly to the chosen use cases, so that people learn by applying AI to real work rather than in isolation.
6. Run pilots, measure and decide
Treat each first use case as a time-limited pilot, typically a few months. Measure results against the success criteria you defined, collect feedback from users, and record what had to change in processes and data.
At the end of each pilot, make a clear decision: scale it, adjust it or stop it. Stopping a pilot that does not deliver is a sign of good management, not failure. Share lessons openly so that the next project starts from a better position.
7. Scale what works and review the strategy regularly
Once a use case proves its value, plan how to extend it to other departments or sites, what support it will need and how it will be maintained. At the same time, revisit the strategy at least once a year. Technology, regulation and your organisation's priorities will change, and the strategy should change with them.
A one-page AI roadmap that shows priorities, pilots, enablers such as data and skills, and governance milestones over the next twelve to eighteen months is often more useful than a long report. It gives leaders a shared reference and makes progress visible.
A practical checklist
- Have we linked AI to three to five clear organisational priorities?
- Do we know the state of our data, skills, technology and culture?
- Have we chosen a small number of measurable use cases?
- Is there an AI use policy and a named owner for each system?
- Do leaders, managers and staff have the skills they need?
- Do we measure pilots and make clear decisions to scale or stop?
Developing the skills to lead AI strategy
BRT Skills offers practical programmes for leaders and teams who are shaping AI strategy in companies and public bodies. Building an Enterprise AI Strategy & Roadmap helps leadership teams turn priorities into a realistic plan, Leading the AI-Enabled Organisation focuses on the leadership and change side, and Data Governance & Data Quality Management builds the data foundations every AI project depends on. You can browse all programmes in our AI training courses section, delivered in London, Dubai, other cities, live online or in-house for your organisation.