Digital Transformation Won’t Fix Your AI Problem

Why “Digital Transformation” Is the Wrong Starting Point for the AI Era

For more than a decade, “digital transformation” has been the default ambition for organisations trying to modernise. It sounds directional. It signals intent. And it suggests progress.

But as generative AI moves from experimentation into the core of enterprise operations, that language is starting to show its limits.

The issue isn’t that transformation is the wrong goal. It’s that treating transformation as the starting point encourages the wrong behaviour—tool-led decisions, fragmented initiatives, and a steady accumulation of pilots that never quite change how the organisation actually works.

The data now makes this clear.

The Spend Is Exploding. The Confidence Isn’t.

According to Gartner, worldwide generative AI spending reached $644 billion in 2025, a 76% increase year on year. Yet nearly 80% of that spend flowed into hardware—servers, PCs, smartphones—rather than into rethinking workflows, decision structures, or operating models.

At the same time, Gartner notes a growing paradox: expectations for GenAI are declining, driven by high proof-of-concept failure rates, even as investment in foundational models accelerates.

In other words, organisations are buying the capability to transform—often at scale—but struggling to turn that capability into sustained operational change.

This is not a technology problem. It’s a starting-point problem.

Transformation Is an Outcome, Not an Activity

Many transformation programmes implicitly assume that introducing new technology causes transformation. The logic is linear: deploy the tool, modernise the process, change the business.

In practice, the causality often runs the other way.

Real transformation tends to emerge when organisations first clarify:

• how work flows across teams,
• where decisions are made,
• how data is created, validated, and reused,
• and which activities genuinely matter to performance.

Only then do tools accelerate change.

IBM’s Institute for Business Value reinforces this in its research on scaling generative AI. Organisations that move beyond isolated use cases and embed AI into core workflows—sales, security, supply chain, operations—see materially higher and more sustainable returns. Those that remain trapped in a “use-case mindset” do not.

Transformation, in other words, is the result of operating model change. Treating it as the activity obscures the work that actually needs to be done.

Reframing the Language Exposes the Gap

When organisations say:

• “We’re undergoing a digital transformation,”
  the more useful question is: what is changing in how work gets done, day to day?
• “This tool will transform the business,”
  the question becomes: relative to which operating model?
• “We need to modernise,”
  the follow-up is: modernise what, exactly—decisions, workflows, accountability, or just interfaces?

These reframes don’t reject the idea of transformation. They expose its limits as a planning construct.

Adoption Is Real—But Mostly Internal

None of this suggests that AI adoption is stalling. Quite the opposite.

Forrester’s 2024 State of Generative AI research showed that over 90% of enterprise AI decision-makers had concrete plans to implement GenAI, with productivity gains cited as the leading objective, followed by innovation, cost efficiency, and revenue growth.

But where those deployments are happening matters.

Both Forrester and IBM found that internal, employee-facing use cases dominate early adoption: knowledge management, employee productivity, software development, internal decision support. External, customer-facing applications tend to follow later—and almost always with humans firmly in the loop.

This pattern is telling. Organisations implicitly recognise that AI delivers value fastest where it augments existing work, rather than attempting to reinvent the business overnight.

That is not transformation in the programme-management sense. It is incremental operating change.

Digital Literacy Shows Up Before the Tool Choice

What separates organisations that progress from pilots to profit isn’t access to models or infrastructure. It’s digital literacy.

Not literacy in the sense of “can people use the tool,” but in the deeper sense: can the organisation reason about its own systems?

Digitally literate organisations:

• understand how decisions propagate,
• recognise where human judgement is essential,
• know which processes can tolerate probabilistic outputs and which cannot,
• and can articulate where automation helps versus where it creates risk.

This is why Forrester and IBM both emphasise human-in-the-loop patterns, especially in early deployments.

Without that foundational understanding, transformation language simply masks confusion.

From Pilots to Scale Requires an Operating Lens

IBM’s agentic AI research is particularly instructive here. It predicts that agentic AI-enabled workflows will grow from around 3% today to 25% by 2026, with “AI-first” organisations outperforming fragmented adopters on revenue, operating profit, and customer satisfaction.

But the same research—and Gartner’s forecasts—also issue a warning. Gartner expects over 40% of agentic AI projects to be cancelled by 2027, citing unclear business value, escalating costs, and inadequate controls.

The difference between those outcomes is not ambition. It’s coherence.

Organisations that treat AI as a layer within a clearly defined operating model can absorb increasing autonomy over time. Those that treat AI as a transformation initiative struggle to govern it, scale it, or justify it.

This is the work Cyferd was built to do. We don’t start with transformation roadmaps or technology assessments. We start by helping organisations understand their own operating logic—how decisions flow, where accountability sits, which processes create value and which create drag. Only then do we introduce AI as an accelerant. That shift in sequence—operating model first, tools second—is what allows our clients to move from pilots to production without the governance crises, cost overruns, or abandonment rates that now define the majority of enterprise AI programmes. The organisations that scale AI successfully aren’t the ones with the biggest budgets. They’re the ones that know what they’re scaling toward.

A Better Starting Question

If “How do we digitally transform?” is the wrong question, what replaces it?

A more productive starting point is simpler and more uncomfortable:

How should this organisation actually operate?

Only once that is clear do questions about AI, automation, orchestration, or transformation become meaningful. Tools stop being the strategy. They become instruments.

Seen this way, digital transformation isn’t abandoned. It’s demoted—from a headline ambition to a by-product of doing the harder, less visible work first.

Language Shapes Outcomes

The persistence of “digital transformation” as a starting point isn’t harmless. Language shapes planning horizons, investment decisions, and success criteria.

As AI becomes ambient, agentic, and increasingly invisible—as Gartner’s “beyond GenAI” framing suggests—the organisations that succeed will not be the ones that transformed hardest, fastest, or loudest.

They will be the ones that understood themselves well enough to change deliberately.

Transformation will still happen.
It just won’t be where it starts.