This is a scenario I encounter regularly. An enterprise has invested meaningfully in AI. They have data scientists. They have cloud infrastructure. They have run several pilots, some of which produced genuinely promising results. And yet the value from AI is consistently below what was promised, what was budgeted for, and what leadership expected.
When I dig into why, the technology is almost never the problem. The models are fine. The data is usable. The individual talent is often excellent. What's missing is the connective tissue: the structure that turns isolated AI capability into organisation-wide AI performance.
That connective tissue has a name. The AI operating model.
An AI operating model defines how an organisation structures its people, processes, governance, and technology to build, deploy, and sustain AI at scale. It is the difference between having AI and being an AI-capable organisation. In my experience, it is the most underinvested dimension of enterprise AI programmes.
Why the operating model question matters more than the technology question
Most enterprise AI conversations focus on the technology: which model, which platform, which vendor, which use case. These are necessary conversations. But they are not sufficient. The organisations that consistently fail to scale AI are not failing because they chose the wrong technology. They are failing because they have not built the organisational infrastructure to use the technology effectively.
For most enterprises the question is no longer whether they can build AI. They can. The question is whether they are structured to scale it. Most are not.
The five pillars of an effective AI operating model
- Structure and governance: How AI capability is organised and how accountability for AI outcomes is distributed across the enterprise.
- Talent and capability: The skills, roles, and learning infrastructure required to build and sustain AI at scale.
- Process integration: How AI systems are embedded into operational processes and decision workflows, not just deployed alongside them.
- AI governance and policy: The policies, controls, and accountability structures that ensure AI is deployed responsibly and in compliance with regulatory requirements.
- Technology and data infrastructure: The platforms, tools, and data architecture that enable AI teams to build, test, deploy, and monitor AI systems efficiently.
The structure question: centralised, federated, or hybrid?
Centralised: one AI function owns everything
A single AI centre of excellence owns all AI capability, from research to deployment. This maximises consistency and prevents duplication. The weakness is distance from the business: centralised AI teams can become disconnected from operational realities, building technically impressive models that don't get adopted.
Federated: AI capability lives in the business units
Each business unit builds its own AI capability. Decisions are fast and context is deep. The cost is fragmentation: duplicated infrastructure, inconsistent standards, and governance gaps that accumulate quietly until they become visible problems.
Hybrid: the model that most mature organisations converge on
A central platform, standards, and governance function supports distributed AI capability in the business units. The centre provides infrastructure, tools, and governance frameworks. The business units provide context, ownership, and adoption. This is harder to design and maintain, but it is the only approach that scales AI without sacrificing governance or business relevance.
Where operating models break down: the four failure patterns
The capability island
AI capability is built in one part of the organisation and never successfully transfers to the business units that need to use it. The capability island syndrome produces impressive demos and disappointing production numbers.
The process bypass
AI systems are deployed without redesigning the processes around them. People use the AI output as optional information rather than an integrated part of their workflow. Value evaporates.
The governance bypass
AI systems are deployed without adequate governance structures. The remediation cost is always higher than the governance investment would have been.
The talent ceiling
An organisation builds strong AI capability among a small group of specialists but fails to develop broader AI literacy across the leadership and operational teams. The specialists become bottlenecks. Adoption stalls.
Where to start: the operating model assessment
Before designing an AI operating model, you need an honest assessment of where you are today. The questions that matter most: Where does AI capability currently sit? Who owns AI outcomes? How are AI systems integrated into operational processes? What is your current governance capability?
You do not need a perfect operating model to start scaling AI. You need a clear-eyed view of where the model is breaking down, and the will to fix it.
Vijay Jaswal is Founder and CEO of Kudo Advisory. Reach him at info@kudoadvisory.com or on LinkedIn.