AI is not SAP: Here’s Why it Needs a Different Playbook

8 June 2026

Mike Klein | Andras Vicsek | Peter Kalmar:

There is a mistake large organizations are making right now on AI adoption. It is understandable. It is well-intentioned. And it may be fatal.

They are treating AI like SAP.

The SAP playbook – steering committee, phased rollout, approved use cases, training modules, compliance tracking – is a rational response to a certain kind of technology change. A technology with a correct way to be used. Fields populated consistently. Processes run in sequence. Behavior converging on a defined standard. In that world, change management has a clear job: move people from wrong to right, close the gap, measure adoption. It works. For SAP, it is exactly right.

AI is not SAP.

There is no canonical way to use a large language model.

Two people in the same role, both highly effective, completely different workflows. The marketing director stress-testing briefs before they reach her team. The operations manager compressing three days of reporting into ninety minutes. Neither is wrong. The right use is the use that fits the work and the person doing it. The moment an organization tries to mandate how people should use AI – approved use cases, standardized workflows, compliance checkpoints – it is not accelerating adoption. It is capping it.

But here is what makes this more than a tactical error. Organizations running the SAP playbook on AI are not just limiting their upside. They are measuring the wrong things entirely – and staying blind to the forces that will determine whether AI makes their organization stronger or more fragile.

Most AI adoption programs are asking one question: how many people are using the tools? It is the least important of the three questions that actually matter.

What Organizations Are Missing

AI is doing three things simultaneously to every organization that deploys it at scale. The first is visible. The other two are not. And the two that are not visible are the ones that will determine whether AI adoption creates organizational capability or organizational fragility.

The first dimension is adoption.

Usage rates, tool uptake, workflow integration. This is where most measurement and change management effort goes. It matters. But it is a lagging indicator, not a leading one – and it tells you almost nothing about whether the organization is actually getting smarter or faster as a result.

The second dimension is acceleration.

AI compresses the distance between information and decision. Options that took days to model can be surfaced in minutes. Analysis that required specialist capacity can be approximated in seconds. For organizations built around fast, distributed decision-making, this is transformative. For organizations still running hierarchical, deliberative approval processes, it creates a new and widening penalty – the system generates insight faster than the organization can act on it.

The real race is not adoption rates. It is whether collective decision-making capability adapts at the speed and with the consistency that change requires.

The third dimension is isolation.

This is the one almost nobody is tracking – and arguably the most consequential.

As people turn to AI for information, advice, and problem-solving they previously sought from colleagues, the informal trust networks that organizations actually run on are quietly shifting.

As AI substitutes for those colleague interactions, those relationships weaken. The issue is not that people stop talking altogether. It is that the pattern of who learns from whom, who validates judgment, and who transfers context begins to change — often unevenly and invisibly. Some key people – the ones who derive energy and influence from being sought out, trusted, and useful to others – may disengage or lose their informal role in the system. And once those networks shift, you cannot reconstruct the baseline. The clock is running now.

Why the Traditional Playbook Fails

The standard AI adoption playbook was built for a different kind of technology change. SAP, Salesforce, an ERP migration: these have a correct way to be used. Change communication in that world has a clear job – define the right behavior, drive people toward it, measure compliance. Close the gap between current and desired.

AI has no canonical use. And applying the compliance model anyway actively suppresses the adaptive, personalized adoption that generates real organizational value.

More fundamentally: the traditional playbook addresses only the first dimension. It has no mechanism for tracking decision velocity or network health. It cannot see what AI is doing to the informal organization. By the time the effects are visible – in disengagement, in slowed decision cycles, in the quiet exit of key connectors – the damage is done.

What a Different Approach Looks Like

Addressing all three dimensions requires three capabilities working together – not sequentially, and not in isolation.

The first is network intelligence.

Organizational Network Analysis maps the informal trust and influence structures that formal org charts never capture. Who do people actually go to for advice, for news, for support? And how do these patterns change as AI becomes a new source of answers, validation, and problem-solving support? 

In the context of AI adoption, there is another question worth asking: who do people turn to for guidance on AI itself?

The answer is revealing. In most organizations, AI expertise clusters in isolated pockets – super-users who are well-connected within their own teams but invisible to the broader organization. Identifying those people, and connecting them deliberately, is itself a high-value intervention. It does not require a full transformation program to generate impact.

Network analysis also establishes the baseline that makes the isolation dimension measurable.

Without a pre-adoption map of trust relationships, you cannot track what AI is eroding, strengthening, or quietly re-routing . Organizations that run that analysis now – before adoption accelerates further – will have data that organizations waiting six months will never be able to reconstruct.

The second capability is communication infrastructure.

Internal communication, properly deployed, is not a broadcast function.

It is an activation layer – the mechanism through which network intelligence becomes organizational movement. The key distinction from traditional change communication is that IC in this context does not mandate behavior. It expands the possibility space.

A story about how a finance analyst reduced her month-end close from three days to four hours does not tell anyone else what to do. It opens a door. Peer-to-peer story networks – activated through identified connectors, not through top-down messaging – are more trusted, more contextually relevant, and more generative than anything a central function can broadcast. IC also plays a specific role in the acceleration dimension: ensuring that people have what they need to make good decisions fast, at the moment of decision rather than in a quarterly town hall.

The third capability is organizational development.

Network intelligence and communication activation create movement. OD creates the conditions that allow movement to hold – leadership capable of delegating decisions to the edge, team cultures with the psychological safety to act on AI-generated insight without waiting for sign-off, and the internal governance structures that give AI adoption the organizational legitimacy it needs to become embedded rather than episodic. Without this layer, networks get activated and revert. Stories circulate without producing capability change. The investment in the first two capabilities loses its compounding effect.

The Convergence

None of these three capabilities works adequately without the other two. Network intelligence without communication activation and OD produces insight that sits in a report. Communication activation without network intelligence is broadcast into a system it cannot see. OD without the diagnostic intelligence from network analysis is built on assumptions about how the organization works rather than evidence about how it actually functions.

What organizations need is not a choice between these approaches. They need a diagnostic that addresses all three dimensions simultaneously – and a development framework that activates all three in sequence.

The organizations that navigate this well will not be the ones with the most sophisticated AI tools or the most comprehensive training programs. They will be the ones that understood, early enough, that AI adoption is not only a technology or training problem, but a network, communication, and decision infrastructure problem – and built their response accordingly. .

That window is open. It will not stay open indefinitely.

+++

Mike Klein is Principal of Changing The Terms and Editor-in-Chief of Strategic Magazine.

Andras Vicsek is founder of Orgmapper.

Peter Kalmar is Founding and Managing Partner of Flow International.