AI Operating Models: Are We Solving AI Risk or Creating New Organisational Friction?

Over the past 18 months, the conversation around AI has evolved dramatically.

The early focus was on experimentation. Could we use AI? Where might it create value? Which tools should we deploy?

Today, many organisations have moved beyond that stage. AI is now becoming embedded in business processes, workflows and decision-making. The challenge is no longer simply adopting AI. The challenge is scaling it without creating chaos.

That was the focus of a recent CIO WaterCooler Digital Boardroom led by Prisilia Mbuy, Director of Service Strategy and AI Enablement at Cloud and Clear, who explored what an effective AI Operating Model looks like as organisations move from pilots and proofs of concept towards enterprise-wide adoption.

One theme stood out above all others.

AI governance is necessary, but it is not sufficient.

Many organisations are investing significant effort in developing policies, controls, risk frameworks and approval processes. These are all important. However, an uncomfortable question is beginning to emerge:

What if, in our efforts to reduce AI risk, we inadvertently increase organisational friction?

We’ve spent years trying to simplify and accelerate work through digital transformation, automation and modern operating models. Yet there is a danger that AI could trigger a new wave of bureaucracy, where every new use case requires additional approvals, committees and oversight before value can be realised.

The most effective AI Operating Models will not be the ones with the thickest governance documents.

They will be the ones that enable organisations to move faster with confidence.

This raises another important shift in thinking.

Historically, governance has often focused on systems, projects and technology ownership. AI does not fit neatly into those categories.

As AI becomes embedded within workflows, recommendations, business decisions and increasingly autonomous agents, a more fundamental question emerges:

Who owns the outcome?

For years, organisations have asked who owns the application, who owns the platform and who owns the budget.

With AI, those questions still matter, but they may no longer be the most important ones.

If a human and an AI system collectively contribute to a customer interaction, an operational decision or a business process, accountability cannot sit with the technology itself. Someone still owns the outcome.

This is where AI Operating Models may need to evolve beyond traditional governance.

Three areas appear increasingly important:

• Clear decision rights around what AI can recommend, escalate or execute.
• Explicit ownership of business outcomes influenced by AI.
• Measurement of value based on business impact, not simply model costs or token consumption.

The arrival of agentic AI only makes these questions more urgent.

As AI moves from generating content to taking actions, organisations need greater clarity around accountability, not less. The more autonomous these systems become, the more clearly defined the organisational boundaries around that autonomy must be.

There is a paradox here.

The more capable AI becomes, the more important human accountability becomes.

Perhaps that is why the most mature AI Operating Models may ultimately be judged on a surprisingly simple criterion:

How much organisational friction do they remove?

Do they help people make better decisions faster?

Do they accelerate delivery while maintaining appropriate controls?

Do they create clarity around ownership and accountability?

Or do they introduce additional layers of process that slow adoption and dilute responsibility?

As organisations continue their journey from AI deployment to AI-enabled transformation, these questions may become more important than the technology itself.

As you build your AI Operating Model, how do you balance governance and accountability without creating the very organisational friction that AI is supposed to eliminate?