The transformation programmes that work are the ones that get governed

Most organisations do not fail at digital transformation because they picked the wrong technology. They fail because nobody owns the outcome after the pilot, the vision was never specific enough to disagree with, and the programme has no way of telling a project that is working from one that is merely still running.

What follows is the structure I use: a way of locating where an organisation actually sits, three pillars that decide where value comes from, and seven phases that put those into a sequence you can staff and fund.

Westerman and his colleagues at MIT and Capgemini found that organisations sort into four groups on two axes — digital capability and leadership capability — and that only one of them compounds. Digital Masters were 26% more profitable than their industry peers and drew 9% more revenue from their physical assets.

The four positions are worth naming, because most executives recognise their own organisation immediately. Beginners have limited, siloed initiatives. Fashionistas have many initiatives and no coordination — the most expensive position of the four. Conservatives have strong governance and are too cautious to use it. Digital Masters have vision, execution and governance at once.

Transformation value arrives through three channels. Customer experience: seamless, personalised, omnichannel engagement. Operational processes: workflows streamlined with automation and real-time analytics. Business models: rethinking how value is delivered so new revenue becomes possible.

The published examples are instructive because they are unglamorous. Burberry improved retention by 20% by integrating digital and physical retail. Codelco cut costs 15% and improved safety with IoT-enabled operations. Nike+ built loyalty through wearables and community rather than through advertising.

Three things are asked of leadership and only three: define a transformative vision the organisation can actually act on, align and motivate people toward it, and govern which initiatives get resourced. Programmes stall on the third far more often than the first.

• Vision — define where the organisation is going in terms specific enough to disagree with.
• Engagement — align and motivate people, at every level, toward that direction.
• Governance — decide what gets funded, what gets stopped, and who answers for each.

Phases one and two, across the first six months, establish vision and leadership capability — appoint someone accountable, create the governance framework, form the cross-functional teams. Phase three, months seven to twelve, invests in the tools. Phase four builds employee engagement and the training that makes the tools usable.

Phase five, months seventeen to twenty, unifies the data foundation and only then extends it with AI. Phase six implements agile and adaptive processes — pilot small, scale what works, collect feedback continuously. Phase seven, the last three months, is sustainability: KPIs, resource reallocation, and monitoring what is emerging next.

Sustainability is not a final checkpoint. It is the continuing discipline of measuring progress, reallocating resources, and renewing the roadmap — which is why it appears as a phase rather than as an afterword.

The framework has held. The environment around it has not. AI adoption has moved from a strategic question to a baseline condition, and the gap between using AI and getting value from it has become the thing worth measuring.

The financial impact is real but modest. Among functions reporting gains, the most common revenue increase was below 5%, and most reported cost savings were below 10%. Adoption has simply moved faster than the organisational systems required to convert it into durable returns.

Which means: scale only where a use case has an accountable owner, measurable value, adoption inside the actual workflow, reliable data, and explicit risk controls. Four of those five are organisational, not technical.

The NIST AI Risk Management Framework gives four functions worth adopting wholesale. Govern: set policy, ownership, risk tolerance and inventory. Map: define context, affected people, intended use and failure modes. Measure: test performance, bias, privacy, security and human oversight. Manage: prioritise risk, monitor in use, respond and improve.

None of this is a reason to move slowly. It is a reason to know, at any moment, which system is doing what, on whose authority, and what would tell you it had stopped working.