The Five Stages of Data & AI Strategy Grief

I’ve been working with a number of clients from one end of the spectrum to the other. Greenfield to Brownfield to spendhugeamountsofmoneyfield!

What I have seen in the latter two is that many of these organisations do not struggle with Data & AI because they lack ambition, it’s mainly down to the fact that they confuse investment with impact.

Spend enough time in boardrooms and a pattern emerges. There is real commitment. Modern platforms have been implemented, data teams have grown, and AI features prominently in strategy documents. The language is confident and forward-looking.

But then someone asks a deceptively simple question: “what has actually improved?”

Not about activity, or architecture, or even hiring. But pure and unadulterated performance!

That question has a habit of changing the tone in the room.

Over time, I have noticed that organisations tend to move through a predictable cycle in answering it. The stages are not dramatic, but they are remarkably consistent. When I did Organisations Psychology, as a base we discussed the more human elements to understand change etc. One of those was the “Five Stages of Grief: denial, anger, bargaining, depression, and acceptance” based on the Kübler-Ross model.

Well, as the week has worn on, I’ve been thinking about this more deeply and if it can be applied to data & AI Strategy. I regretfully state it can! I’ve changed the five stages slightly and here is my attempt:

Stage 1: Denial – “We’re already data-driven.”

At the beginning, confidence runs high. Dashboards are everywhere, reporting is faster than it used to be. The data platform looks sophisticated and there is a roadmap filled with sensible terminology about modernisation and enablement.

On the surface, this feels like progress.

What is rarely examined closely is whether critical business decisions are being made differently and measurably better as a result. Not whether reports exist, but whether decisions around pricing, risk, allocation, or customer action have improved in ways that show up financially.

When that distinction is made explicit, the room often goes quiet.

Stage 2: Frustration – “Why aren’t we seeing ROI?”

Twelve to eighteen months later, the scrutiny sharpens. The CFO wants evidence of return and the board wants tangible impact. Business leaders complain about data quality and data teams point out that requirements were never clearly defined.

The thing is that the blame game starts and it circulates efficiently. Technology is criticised, engagement is questioned, sponsorship is deemed insufficient and good old governance is described as either too heavy or too weak.

Everyone is working hard. That is not the issue.

The issue is structural. Most strategies were built around capability i.e. platforms, tooling, organisational charts rather than around clearly defined commercial decisions with explicit ownership and measurable outcomes. Capability is visible and fundable, but value requires design discipline.

Stage 3: Overcompensation – “AI will unlock it.”

Of course, there is a silver bullet, right? HA! I got you there didn’t I! When better reporting fails to materially shift performance, attention turns to something more advanced. Predictive models are commissioned, automation initiatives are launched and GenAI pilots begin. Because that will make everything right, the age of sitting back and letting it all happen via an expensive algorithm is the way!

The reasoning appears sound. If information did not change behaviour, perhaps intelligence will.

Yet a fundamental constraint remains. You cannot optimise a decision that no one formally owns. You cannot automate a process that is not consistently measured, and you cannot scale insight if acting on it is optional.

Introducing advanced AI into an ambiguous operating model does not resolve the ambiguity, it only accelerates it.

Stage 4: Fatigue – “Let’s be pragmatic.”

After enough pilots and partial implementations, enthusiasm begins to cool. This should be very familiar, where budgets tighten, roadmaps narrow and language shifts from transformation to prioritisation.

There is talk of consolidation and realism. Where was reality before this? Sitting in the corner of the room sipping soda water while the previous stages are all lording it over Krystal!

What is really happening is a loss of conviction, not in the potential of data or AI, but in the organisation’s ability to convert them into measurable performance improvement. Without a clear line between insight and accountable action, investment begins to feel discretionary.

This is where many strategies stall. The infrastructure remains in place, the teams remain employed and the narrative simply moves on. Until someone ultimately pulls the plug! it happens!

Stage 5: Clarity – “This is about decisions.”

A smaller group of organisations reach a different conclusion and stop starting with data. Instead, they begin with the handful of decisions that genuinely move financial, risk, or customer outcomes.

They identify who owns those decisions. They define what “better” means in measurable, commercial terms. They track behavioural change, not dashboard usage. They design data and AI capabilities specifically to improve those decision loops.

At this point, the results look different, even when the underlying technology is largely the same.

The shift is not technological, it is organisational and the operating model tunes itself. Accountability becomes explicit, outcomes are defined upfront and investment is tied to performance rather than activity or outputs.

It’s Your Prerogative!

Data & AI strategy rarely falters because the technology is immature. It falters because organisations mistake motion for momentum and capability for value.

If there is a way to avoid the grief cycle, it is straightforward, though not easy. Start with the decisions that materially affect performance, assign ownership for improving them, tie that improvement to money, risk, or customer outcomes. Then build the data and AI capability around that spine.

Anything else may look progressive.

It simply will not change the numbers.

About the Author

Samir Sharma is a senior data, analytics, and enterprise applications leader with over 20 years’ experience building and operating data platforms and analytical capabilities within regulated and complex organisations. He specialises in helping executive leadership teams operationalise data and AI strategy across business processes, decision-making, and service delivery. He is the author of The Strategy Canvas: A Field Guide for Data & AI — Closing the Strategy-Execution Gap.

If your organisation is struggling to translate data and AI ambition into measurable business outcomes, contact Samir to discuss how the Data & AI Strategy Canvas framework can help close the gap between strategy and execution.