The Most Important Part of Ai Transformation Isn't AI

The Most Important Part of AI Transformation Isn’t AI

Before AI can transform your business, it needs to know what your business knows.

Which comes first: AI readiness or organisational readiness?

It's becoming something of a corporate chicken-and-egg problem.

Businesses are understandably racing to work out how AI can improve productivity, decision-making, customer experience and almost every other part of the organisation.

But there's an uncomfortable question underneath all of this.

What exactly are we giving the AI to work with?

Imagine recruiting an extraordinarily capable new member of your leadership team.

On their first morning you give them six different versions of the company strategy.

Finance gives them one definition of an active customer. Sales gives them another.

There are three price lists in circulation, nobody is entirely certain which is current, and half the information they need sits inside people's heads.

One department tells them a process works one way. Another tells them it works differently.

Then, at 10am, you ask them for a detailed analysis of the business and expect the answer to be accurate.

Nobody would call that an employee-performance problem.

We would call it an organisational-alignment problem.

Yet this is remarkably close to what many businesses are now asking AI to do.

AI may be exposing a problem that was already there

The difficulty isn't necessarily that organisations don't have enough information.

Most have enormous amounts of it.

The problem is that information has accumulated rather than been designed.

Documents have been duplicated. Spreadsheets have multiplied. Systems have been added. Teams have developed their own terminology. Knowledge sits in inboxes, shared drives, CRMs, databases and people's heads.

This worked—imperfectly—when humans acted as the integration layer.

Someone usually knew which spreadsheet was current. Someone remembered why the number had changed. Someone knew that the process document hadn't been updated since the new system went live.

AI changes the economics of that mess.

Suddenly we want software to search, interpret, connect and act upon organisational knowledge at enormous speed.

And computers are rather less forgiving of organisational folklore.

The evidence suggests businesses are discovering this quickly. Gartner reported that 63% of organisations either do not have—or are unsure whether they have—the right data-management practices for AI. It predicted that through 2026, 60% of AI projects unsupported by AI-ready data would be abandoned. ¹

IBM's 2025 study of 2,000 CEOs found that 72% believed proprietary data was key to unlocking generative AI's value and 68% considered an integrated enterprise-wide data architecture critical for cross-functional collaboration. Yet 50% also said the speed of technology investment had resulted in disconnected technology inside their organisations. ²

The AI problem, in other words, may partly be a problem AI has revealed.

Teams have faced the same problem for years

There is an interesting parallel here with what we already know about effective teams.

A high-performing team does not require every member to know everything.

That would be impossible.

But team members do need enough of a shared understanding to work together effectively.

Researchers call these shared mental models: common understandings of the task, the environment, priorities, roles and how the team operates. Research has consistently linked stronger shared mental models with team processes and performance. ^3,4

There is an even more interesting concept called a transactive memory system.

In simple terms, a good team develops a collective understanding of who knows what.

One person doesn't need to remember everything because they know who holds the expertise, they trust that expertise, and the team knows how to retrieve and coordinate it.

A 2026 meta-analysis covering 44 studies found a moderate positive relationship between transactive memory systems and team outcomes. ⁵

That begins to sound remarkably similar to the challenge businesses face when organising information for AI.

The objective isn't necessarily to create one enormous database containing everything.

It is to establish something more important:

Where does the truth live?

Which system owns customer information?

Which record contains the authoritative product information?

Where is pricing controlled?

Who is allowed to change it?

What happens when it changes?

Which version should every other system—and increasingly every AI agent—rely upon?

This is the logic behind a Single Source of Truth.

And viewed through the lens of teams, an SSOT isn't simply an IT architecture.

It is a form of organisational alignment.

A Single Source of Truth is a behaviour, not just a database

This is an important distinction.

Creating a Single Source of Truth does not mean moving every piece of information in the company into one enormous system.

It means creating a discipline around information.

Something is created once.

There is an agreed owner.

There is an authoritative location.

Other systems reference it rather than creating alternative truths.

When something changes, it changes at source.

That sounds technical. But achieving it is profoundly human.

People have to stop keeping their own versions.

Departments have to agree definitions.

Ownership has to become explicit.

Teams have to trust information created elsewhere.

Someone has to be accountable for accuracy.

And people have to accept that convenience—"I'll just make my own spreadsheet"—sometimes creates complexity for everybody else.

This may explain why data governance is increasingly being discussed as a cultural issue rather than merely a technology issue.

In September 2026, Gartner reported that among surveyed data and analytics leaders, cultural resistance outweighed funding constraints as the primary reason governance initiatives failed: 60% versus 40%. Gartner's argument was particularly telling: organisations need not just AI-ready data, but AI-ready stakeholders who understand and participate in creating trusted information. ⁶

That sounds much less like an IT project.

It sounds like team development.

So which does come first?

The tempting answer is data.

Sort out the information, create the Single Source of Truth, connect the AI and off you go.

Reality is probably more interesting.

It is a feedback loop.

AI requires organisational alignment. But the process of becoming AI-ready also forces organisational alignment.

The moment you ask an AI system to answer meaningful questions across the organisation, uncomfortable questions start appearing.

Which number is correct?

Which document should it trust?

Who owns this information?

Why do these two departments define the same thing differently?

Why is this stored three times?

Why isn't this captured at all?

AI creates pressure to answer questions that organisations have sometimes been able to avoid for years.

So perhaps the chicken-and-egg question has the wrong answer.

Neither comes first.

They develop together.

The pursuit of AI readiness forces better information discipline. Better information discipline makes AI dramatically more useful. Greater AI capability then exposes the next layer of organisational inconsistency.

And the cycle continues.

From a shared mental model to a machine-readable one

Perhaps this is where our understanding of teams needs to evolve.

For years we have talked about alignment, shared purpose, clear roles, good communication, effective handovers, common language and trusted information as characteristics of strong human teams.

None of those requirements disappears when AI arrives.

They become more important.

The difference is that part of the team can now be digital.

Researchers studying human-AI teaming are already applying concepts such as shared mental models and team cognition to relationships between people and AI systems. The underlying question is increasingly similar to one we have always asked of human teams: do the participants have enough common understanding to coordinate effectively? ⁷

That suggests a different way for leaders to think about AI readiness.

It isn't simply:

Which AI tools should we buy?

It is also:

Does our organisation have a sufficiently aligned understanding of itself for AI to work alongside us?

You could describe a good Single Source of Truth as a kind of machine-readable organisational alignment.

It tells both people and technology what the organisation currently believes to be true.

And once that foundation exists, something important changes.

AI stops being merely a clever tool sitting outside the organisation.

It can begin working with the organisation's actual knowledge.

The next generation of teams

This is why AI shouldn't only be a conversation for technology departments.

It is becoming a conversation about how organisations work.

Data ownership.

Decision rights.

Trust.

Communication.

Knowledge sharing.

Breaking silos.

Clear handovers.

Common language.

Accountability.

Continuous learning.

These are not new challenges.

Anyone who has worked with teams will recognise every one of them.

What is new is that we are beginning to introduce intelligent technology directly into that system.

And perhaps that's the bigger AI opportunity.

Not simply making individual employees faster.

But creating organisations in which people, information and intelligent systems are sufficiently aligned that they can operate as one connected system.

Because before AI can understand your business, there is a more fundamental question.

Does your business understand itself?

Research references

1. Gartner. Lack of AI-Ready Data Puts AI Projects at Risk. 26 February 2025.

2. IBM. IBM Study: CEOs Double Down on AI While Navigating Enterprise Hurdles. 6 May 2025.

3. Leslie A. DeChurch and Jessica R. Mesmer-Magnus. Measuring Shared Team Mental Models: A Meta-Analysis. 2010.

4. Leslie A. DeChurch and Jessica R. Mesmer-Magnus. The Cognitive Underpinnings of Effective Teamwork: A Meta-Analysis. 2010.

5. Crystal M. Fausett, Joseph R. Keebler, Elizabeth H. Lazzara, Megan Gregory and Elizabeth Blickensderfer. Measurement Matters: A Meta-Analytic Examination of Transactive Memory Systems and Team Outcomes. 27 April 2026.

6. Gartner. Gartner Predicts 60% of Organizations That Ignore Data Governance Culture Challenges Will Fail to Govern AI Successfully by 2027. 21 September 2026.

7. Shaida Kargarnovin and colleagues. From testbeds to high-stakes work: a review of Human-AI teaming domains and teaming factors. 7 May 2026.