In 2012, McKinsey declared digital transformation the defining business challenge of the decade. Companies hired Chief Digital Officers, stood up Centers of Excellence, ran 90-day sprints, and called it transformation. By 2020, the data was in: roughly 70% of those initiatives had failed to deliver on their objectives.
Now listen to how people are talking about AI in 2026.
"It's an existential threat if you don't move fast." "Start with strategy, not technology." "Culture eats AI for breakfast."
We have been here before. The words have changed. The pattern hasn't.
"The companies that got digital transformation right didn't do it by buying more software. They changed how decisions get made."
What Actually Killed Digital Transformation
The 70% failure rate wasn't caused by bad technology. The enterprise software that powered digital transformation was largely excellent. The technology worked. The transformations didn't.
Three things killed most DT initiatives — and all three are human and organisational, not technical.
They digitised existing processes instead of reimagining them. The most common failure mode was taking an analogue process, making it digital, and calling it transformation. The underlying process — with all its inefficiencies and outdated assumptions — remained untouched.
They changed the technology without changing the operating model. The CDO reported into IT. The transformation team sat outside the business. The core decision-making structures were never touched. Companies bought the technology of transformation without buying the organisational change that gives it meaning.
They measured the wrong things. The vanity metrics proliferated: number of employees trained, apps deployed, processes automated. The real question — "which decisions are being made faster, better, cheaper?" — was rarely asked and almost never answered.
The AI Transformation Parallel
The parallels to today's AI moment are uncomfortable.
| The pattern | Digital Transformation (2012–2020) | AI Transformation (2023–present) |
|---|---|---|
| The structural fix | Hire a CDO. Stand up a Center of Excellence. | Hire a Head of AI. Stand up an AI lab. |
| The measurement trap | Processes digitised. Apps deployed. Employees trained. | AI tools deployed. POCs completed. Prompts written. |
| The process mistake | Digitise existing processes without questioning them. | Automate existing workflows without questioning them. |
| The org failure | DT team outside the business. Business waits to be served. | AI team outside the business. Business waits to be served. |
| The ROI confusion | Spent €2M. Can't name a single changed decision. | Spent €2M. Can't name a single changed decision. |
The last row is not hypothetical. A CFO told me those exact words about her company's AI spend earlier this year.
Three Lessons DT Teaches AI
The One Difference That Makes It Harder
If the patterns are the same, why does it matter that we're talking about AI and not digital transformation?
Because the speed is approximately three times higher, and the margin for error is correspondingly smaller.
Digital transformation played out over a decade. Companies that made the wrong structural decisions in 2012 had until 2018 to course-correct. The pace of AI capability development compresses that same decision cycle into two or three years.
You have less time to make the same mistakes. The pattern is identical. The urgency is not.
The Transformation Layer
A weekly 300-word essay on AI transformation, infrastructure, and what executives get wrong.