Explainer · Part 1 of 6
What AI transformation is, and what it is not
Adoption is near-universal; value is not. AI transformation is the redesign of work, decision rights and the operating model so that AI produces measured value at scale.
AI transformation is the sustained redesign of work, decision rights and the operating model around AI, governed across its lifecycle and held accountable for measured value at enterprise scale. It is not the adoption of AI tools.
The distinction matters because adoption is now almost universal while value is not. In McKinsey's 2025 global survey, 88% of organisations use AI in at least one function, yet only 7% report it fully scaled and about 6% attribute 5% or more of EBIT to it (IDM summary).
A ladder of older concepts
The term is used loosely, so it helps to place it on a ladder of older ideas.
| Concept | What changes | Unit of change | Typical evidence of success |
|---|---|---|---|
| Digitisation | Analogue information becomes digital | A record or form | Data captured digitally |
| Digitalisation | An existing process runs on digital technology | A process | Cycle time, cost per transaction |
| Digital transformation | The organisation's properties change through information, computing, communication and connectivity technologies | The organisation | New capabilities, business model shifts |
| AI adoption | AI tools are made available and used | A user, team or use case | Licences issued, active users, pilots run |
| AI transformation | Work, decision rights and the operating model are redesigned around AI and governed for value | Workflows, roles and the operating model | Measured outcomes against a baseline, sustained at scale |
Vial (2019) defines digital transformation as "a process that aims to improve an entity by triggering significant changes to its properties" through information, computing, communication and connectivity technologies. AI transformation inherits that definition but differs in four properties that change how it must be managed.
Four properties that make AI different
- AI is a general-purpose technology. General-purpose technologies spread across sectors and need complementary invention before they pay off (Bresnahan and Trajtenberg, 1995). For large language models, Eloundou et al. estimate that about four in five US workers could have at least 10% of their tasks affected, and about 19% at least half. The change lands in work design across the organisation, not in one system.
- Outputs are probabilistic. A digitised process produces the same result every time; an AI system does not. Assurance cannot end at go-live: evaluation, monitoring and re-validation continue for as long as the system runs.
- AI now acts, not only advises. Agentic systems take actions, call tools and hand work to people and other agents. Decision rights, accountability and human oversight become design choices, not afterthoughts.
- Value depends on complementary investment. Returns arrive only after organisations invest in new processes, skills and data, much of it intangible and invisible in the accounts (Brynjolfsson, Rock and Syverson, 2021).
Three tests
Three tests separate transformation from adoption.
- Scope. The change reaches the enterprise, not a pilot.
- Mechanism. Work is redesigned, not a tool layered on unchanged work.
- Accountability. Value is measured against a baseline after deployment, and governance runs for the life of each system.
An initiative that fails any of the three is adoption, however large its budget.
Sources
- Bresnahan, T. and Trajtenberg, M. (1995). General purpose technologies: "Engines of growth"? Journal of Econometrics.
- Brynjolfsson, E., Rock, D. and Syverson, C. (2021). The productivity J-curve. American Economic Journal: Macroeconomics.
- Eloundou, T., Manning, S., Mishkin, P. and Rock, D. (2023). GPTs are GPTs. arXiv 2303.10130.
- Vial, G. (2019). Understanding digital transformation: a review and a research agenda. Journal of Strategic Information Systems.
- IDM: McKinsey State of AI 2025 (secondary report of the survey).