AI Governance Belongs Inside the Transformation, Not Bolted On After
A pattern shows up in enterprise after enterprise right now. The ERP or platform transformation has its own steering committee, its own stage gates, its own risk register. The AI initiative, often launched around the same time, has none of that. It answers to a different sponsor, reports through a different channel, and gets evaluated on a different timeline. Two governance structures, one organization, and no shared picture of risk.
How the split happens
It rarely happens on purpose. AI pilots often start small, inside a single department, without the overhead of a formal program. That is reasonable for a proof of concept. The problem is that pilots succeed and quietly graduate into production without ever being folded into the governance the rest of the transformation already has in place.
By the time anyone notices, the AI initiative is touching the same data, the same workflows, and often the same users as the core transformation, but it was never assessed against the same standards for risk, readiness, or benefit realization.
What running two operating models actually costs
The immediate risk is technical. AI systems built on top of a data environment that is mid transformation inherit whatever is unstable about that environment, without anyone tracking the dependency. The less obvious risk is organizational: two governance tracks compete for the same executive attention, the same change management bandwidth, and the same frontline capacity to absorb new ways of working. Employees end up navigating two separate change efforts that were never designed to be experienced together.
Sponsors are often the last to see this, because each initiative reports its own status independently, and each one can look healthy in isolation while the combined load on the organization is unsustainable.
The accountability picture makes the case even more directly. Research this year found that fewer than a third of organizations say their CEO takes direct responsibility for AI governance oversight, and only a small fraction say their board holds that accountability. When the AI initiative sits outside the transformation's governance structure, it frequently sits outside anyone's clear accountability at all.
What integration actually requires
Folding AI governance into the existing transformation structure does not mean slowing AI adoption down to match the pace of a larger program. It means the same steering committee sees both, the same risk register tracks both, and the same readiness assessment that applies to the ERP rollout applies to how AI tools are being deployed against that same data and those same workflows.
In practice, this is often a matter of a single agenda item, not a parallel bureaucracy: a standing question at every governance review about how the AI initiative's risk profile has changed, asked with the same rigor as every other workstream.
The window for doing this well is closing
Organizations early in a transformation still have the option to build AI governance in from the start. Organizations further along are discovering that retrofitting oversight onto an AI initiative that has already scaled past a department pilot is considerably harder than building it in from day one. The earlier the two get treated as one program instead of two, the less expensive that alignment is.
Figures cited on AI governance accountability are drawn from 2026 industry research, including Liminal's enterprise AI governance research. Provided for context, not as legal or compliance advice.
Get in touch if AI adoption is running alongside a larger transformation in your organization and you want an independent read on how well the two are actually aligned.