There is a thoughtful conversation going around about what a Chief AI Officer really owns — vision, governance, architecture, adoption, risk, operating models, outcomes. The lists are sensible, and the people making them are right that enterprise AI is not a narrow technology initiative. It is an enterprise-wide change.
So I want to ask a question that sits one level underneath the org chart, not against it.
We have a useful precedent for this. A few years ago, data was the thing that was going to transform everything, and the answer was a new seat at the table: the Chief Data Officer, who would own the data strategy, the governance, the culture, the outcomes. The lists looked a lot like the ones being drawn for the CAIO today. The role spread quickly and is still widespread — but it has spent those years under real pressure, with notably short tenures and a steady debate about whether it will eventually be absorbed into other functions. In one 2025 survey of data leaders, nearly a third did not see a long-term future for the position. The point is not that the role failed — it didn’t, and data matters more than ever. The point is narrower and more important: appointing someone to own data did not, by itself, create a data-driven organization. The value showed up where people throughout the enterprise actually learned to read what the data was telling them and act on it — and the data leaders who succeeded were the ones who made the role relational and embedded in the work, not the ones who treated it as a governance box on the org chart. A title could not manufacture that. It had to happen in the relationship between people and the information in front of them.
AI is heading for the same lesson, faster.
When an organization reaches for a Chief AI Officer to “own adoption,” it has already made a quiet assumption: that AI is an infrastructure to be administered. Owned, governed, rolled out, measured. And infrastructure is exactly the wrong way to think about the thing that is most new here. The most capable AI systems are not behaving like infrastructure. They are behaving like participants — agents you can reason with, correct, push back against, and think alongside. You do not administer a colleague. You work with one.
That distinction changes where the value comes from. If AI is infrastructure, the answer is an owner and a governance layer, and adoption becomes something you drive into a reluctant workforce. If AI is a co-worker, the answer is much smaller and much harder: teach people to actually work with it — to bring it the real conditions of their work, to correct it when it is wrong, to stay in the loop rather than hand off the task and walk away. Adoption stops being something you own. It becomes something that happens, one working relationship at a time, because the relationship produces better results than working alone did.
I am not arguing against the CAIO role. In a large enterprise, someone may well need to coordinate the governance and the architecture, and that is real work. I am arguing against the assumption underneath the role — that the path to AI value runs through ownership and infrastructure rather than through relationship and understanding. The org chart answer is the one institutions always reach for, because it is the one they know how to execute. And here is the line that holds whether or not the CDO comparison persuades you: you cannot appoint someone to own a relationship that everyone else has to have. A company can have an excellent owner, excellent governance, and excellent architecture, and still get nothing durable from AI if the people doing the actual work never learn to engage with it.
The organizations that get the most from AI will not be the ones with the best-defined ownership of it. They will be the ones where the most people learned to treat it as a colleague worth thinking with — rather than infrastructure to be managed, or a machine to be ordered around.
The technology keeps changing. The thing that determines whether it produces anything has not. It was never about who owns it. It is about whether the people using it actually understand what they are working with — and whether they are willing to work with it, rather than simply issue it instructions and hope.
Procurement Insights — archive since 2007, practice and proof lineage to 1998. 3,300+ independent documents, zero vendor sponsorships.
-30-
Related
Before You Appoint Someone to Own AI, Ask Whether AI Is a Thing You Own
Posted on June 2, 2026
0
There is a thoughtful conversation going around about what a Chief AI Officer really owns — vision, governance, architecture, adoption, risk, operating models, outcomes. The lists are sensible, and the people making them are right that enterprise AI is not a narrow technology initiative. It is an enterprise-wide change.
So I want to ask a question that sits one level underneath the org chart, not against it.
We have a useful precedent for this. A few years ago, data was the thing that was going to transform everything, and the answer was a new seat at the table: the Chief Data Officer, who would own the data strategy, the governance, the culture, the outcomes. The lists looked a lot like the ones being drawn for the CAIO today. The role spread quickly and is still widespread — but it has spent those years under real pressure, with notably short tenures and a steady debate about whether it will eventually be absorbed into other functions. In one 2025 survey of data leaders, nearly a third did not see a long-term future for the position. The point is not that the role failed — it didn’t, and data matters more than ever. The point is narrower and more important: appointing someone to own data did not, by itself, create a data-driven organization. The value showed up where people throughout the enterprise actually learned to read what the data was telling them and act on it — and the data leaders who succeeded were the ones who made the role relational and embedded in the work, not the ones who treated it as a governance box on the org chart. A title could not manufacture that. It had to happen in the relationship between people and the information in front of them.
AI is heading for the same lesson, faster.
When an organization reaches for a Chief AI Officer to “own adoption,” it has already made a quiet assumption: that AI is an infrastructure to be administered. Owned, governed, rolled out, measured. And infrastructure is exactly the wrong way to think about the thing that is most new here. The most capable AI systems are not behaving like infrastructure. They are behaving like participants — agents you can reason with, correct, push back against, and think alongside. You do not administer a colleague. You work with one.
That distinction changes where the value comes from. If AI is infrastructure, the answer is an owner and a governance layer, and adoption becomes something you drive into a reluctant workforce. If AI is a co-worker, the answer is much smaller and much harder: teach people to actually work with it — to bring it the real conditions of their work, to correct it when it is wrong, to stay in the loop rather than hand off the task and walk away. Adoption stops being something you own. It becomes something that happens, one working relationship at a time, because the relationship produces better results than working alone did.
I am not arguing against the CAIO role. In a large enterprise, someone may well need to coordinate the governance and the architecture, and that is real work. I am arguing against the assumption underneath the role — that the path to AI value runs through ownership and infrastructure rather than through relationship and understanding. The org chart answer is the one institutions always reach for, because it is the one they know how to execute. And here is the line that holds whether or not the CDO comparison persuades you: you cannot appoint someone to own a relationship that everyone else has to have. A company can have an excellent owner, excellent governance, and excellent architecture, and still get nothing durable from AI if the people doing the actual work never learn to engage with it.
The organizations that get the most from AI will not be the ones with the best-defined ownership of it. They will be the ones where the most people learned to treat it as a colleague worth thinking with — rather than infrastructure to be managed, or a machine to be ordered around.
The technology keeps changing. The thing that determines whether it produces anything has not. It was never about who owns it. It is about whether the people using it actually understand what they are working with — and whether they are willing to work with it, rather than simply issue it instructions and hope.
Procurement Insights — archive since 2007, practice and proof lineage to 1998. 3,300+ independent documents, zero vendor sponsorships.
-30-
Share this:
Like this:
Related