Making AI Transformation Stick: A Participatory Approach

The approach

The approach followed the four stages that now form my DATA framework, moving the organisation from fragmented AI use to a governed, adopted strategy.

• Discover: evidence was gathered from two directions at once. A leadership workshop surfaced a longlist of around 15 activities across the business that added genuine value and could be AI-augmented, from research to reporting to meeting summarisation. In parallel, an employee survey captured the ground truth: how people were actually using AI day to day, how often, and, critically, which tools they were using that leadership didn't know about.

• Analyse: the longlisted activities were plotted on a Value vs Volume matrix, isolating the six to seven activities that were both high-value and high-frequency, the true candidates for AI acceleration. Available AI tools were then reviewed against these specific priority activities, assessing fit and gaps. Cross-referencing this against the survey data revealed the scale of the shadow-AI problem: the tool most people were already using daily (ChatGPT) was not the tool best suited to the organisation's highest-value task, and the licensed tool (Copilot) was underused relative to its capability.

• Tailor: from this diagnosis, a bespoke response was designed: a set of AI principles built around ethics, data protection and innovation (summarised in a working rule: use AI on high-volume tasks to free up capacity for high-value work), a tool recommendation matched to the priority activities identified, and a structured prompting method the team could apply to that top-priority activity, trends research.

• Activate: this was delivered through a hands-on AI Lab training session, run in small groups, underpinned by change management principles to help new behaviours stick rather than fade after a single workshop. Deliberately, the session tested Copilot, Claude and Gemini side by side on a live trends-research task, not ChatGPT, since the survey had already shown it needed no further promotion. Results were compared live and discussed as a group, turning the strategy from a document into a shared, practical reference point the team could return to.

What changed as a result?

The engagement delivered a clear AI strategy framework, a validated set of tool recommendations grounded in real usage data rather than assumption, and a trained team equipped with a repeatable prompting method for one of the organisation's highest-value activities. By building the training on change management fundamentals, not just teaching a tool but embedding a new way of working, the prompting techniques introduced in the session continued to be used by the team afterwards, a sign the approach had genuinely taken hold rather than simply been presented.

FAQ

How do you decide which activities to prioritise for AI adoption?

Candidate activities are plotted on a Value vs Volume matrix, identifying tasks that are both high-value and high-frequency, the areas where AI acceleration has the most impact for the least disruption.

How do you address ungoverned or "shadow" AI use in an organisation?

It starts with measuring it. An honest employee survey on actual tool use, not assumed use, reveals the real gap between licensed tools and what people have adopted informally, which is essential before recommending any change.

Why exclude the most popular AI tool from training?

If a tool is already in daily use, training time is better spent building fluency in higher-value or under-adopted alternatives. Tool recommendations are matched to the task, not to popularity.