Should IT Own AI? The Missing Role in Your Company
No, IT should not own AI. IT should own the tools, the data access and the security. The business should own the results. Between the two you need a product function: one person whose job is to turn a business problem into a working AI workflow that people actually use.
Most companies of 25 to 500 people do not have that person. So AI lands on the desk of whoever seems closest to technology, usually the IT lead or a fractional CTO, and stalls there. This post explains why that happens, what the missing role does, and how to set it up without hiring a new department.
Why does AI keep landing on IT's desk?
Because AI looks like software, and software has always belonged to IT. When a CEO says "we need to do something with AI," the request flows to the person who manages licenses, laptops and logins. That feels sensible. It is also how AI becomes a tool rollout instead of a change in how work gets done.
Even large companies have not settled this. KPMG's Q2 2026 AI Pulse survey of leaders at $1 billion-plus companies found that accountability for AI-informed business decisions is split, with only 14% placing it with a business unit leader. In a Pearl Meyer survey of executives and board members, C-suite respondents spread accountability for AI strategy across the C-suite, the level below it, individual business leaders and functional heads. If big companies with AI teams cannot agree, a 120-person company with a two-person IT team will default to whoever is nearest.
What goes wrong when IT owns AI?
Nothing goes wrong with IT. The problem is the job description. IT is measured on uptime, security and cost. AI value is measured on cycle time, margin, revenue and hours given back to the team. Those are different scoreboards, and people play to the scoreboard they are given.
When IT owns AI, you tend to see the same pattern:
- Tools get licensed, not used. Everyone gets access to an assistant. A few people use it daily. Most open it twice.
- Projects start from the tool, not the problem. The question becomes "what can we do with this?" instead of "what is slowing down our quoting process?"
- Workflows stay the same. AI gets bolted onto an old process, so the time saved disappears back into the day.
- Nobody owns the number. IT can report adoption. Nobody reports whether proposals go out faster or margins moved.
That last point is where the money is. McKinsey's 2026 State of AI report found that nearly three-quarters of high performers report fundamentally redesigning workflows because of their AI use, compared with about a quarter of other respondents. Redesigning a workflow is not an IT task. It needs someone who understands the work, the people and the tools at the same time.
What is the missing product function?
A product function, in this context, is the role that decides what an AI solution should do, for whom, and how you will know it worked. It is borrowed from software companies, where a product manager sits between customers and engineers. In your company, the "customers" are your own teams and the "engineers" are your tools, vendors and IT.
MIT's research on why generative AI pilots stall points the same way. As Fortune reported, the core problem is a learning gap for both tools and organizations, not model quality, and a key success factor is putting line managers, not just central AI labs, in charge of driving adoption. That is a product job: someone close to the work, accountable for whether the thing gets used.
Who owns what: business, product and IT?
The simplest way to stop the ping-pong between departments is to write down who owns what. Here is a split that works for most mid-market companies:
- The business leader owns the outcome. The head of sales, finance or operations names the problem, owns the metric and changes how the team works. If the number does not move, it is their miss.
- The product owner owns the solution. They map the current workflow, decide what AI should do in it, choose whether to configure, buy or build, run the pilot, and drive adoption until it sticks.
- IT owns the foundation. Security, data access, integrations, vendor review and support. IT says how something can be done safely, not whether it is worth doing.
- The CEO owns the portfolio. Which problems get attention this quarter, which get dropped, and who is accountable for each.
Notice that IT is still essential. It just stops being the default owner of things it cannot control, like whether a sales team changes its habits.
Who should play the AI product role in a 25-500 person company?
You probably do not need to hire for it on day one. You need to assign it. There are three common options, and each has a trade-off.
- An operations leader with curiosity. Often the best fit. They already think in processes and know where the friction is. The risk is that AI becomes a side project on top of a full-time job, so carve out real time, at least a day a week.
- A rising manager from the business. Someone who already uses AI on their own work and is respected by peers. Good for adoption, because people trust one of their own. They will need support on vendor and data questions.
- An outside partner or fractional product lead. Useful when nobody inside has the time or the range. The test is whether they leave capability behind or just a set of tools.
What matters more than the title is the mandate. The product owner must report to someone senior, ideally you, and must be able to say no to projects that do not tie to a business number.
What does an AI product owner actually do each week?
The work is practical and repetitive, which is why it gets skipped. A typical week looks like this:
- Sits with the team doing the work. Watches how a quote, a month-end close or a support ticket really moves, step by step.
- Writes a one-page brief per use case. The problem, the metric, the workflow before and after, and what "done" looks like.
- Picks the lightest option that works. Usually switching on AI in a tool you already pay for before buying anything new.
- Runs short pilots with a stop date. Two to six weeks, with a clear call at the end: scale it, fix it or kill it.
- Tracks use, not just access. Who uses it, how often, and what they still do by hand.
If that list sounds like change management as much as technology, that is the point. AI that is not used is not AI.
How do you set this up in the next 90 days?
You can put the product function in place within a quarter without a reorganization. Here is a simple sequence:
- Week 1: Write down who owns what. Use the four-way split above. Share it with your leadership team and IT so everyone knows their part.
- Weeks 1-2: Name the product owner. Give them a mandate in writing, protected time and a direct line to you.
- Weeks 2-4: Pick one or two problems. Choose problems tied to a number you already report. If you need a method, see our guide on how a CEO should prioritize AI projects.
- Weeks 4-10: Run the first pilot inside the real workflow. Not beside it. The business leader changes the process, the product owner runs the pilot, IT keeps it safe.
- Weeks 10-12: Review and decide. Did the number move? Are people using it without being reminded? Scale, fix or stop, then pick the next problem.
This rhythm is what moves a company from running scattered pilots to having AI inside core work. Our post on the AI maturity model explains why that jump, from Experimenting to Operating, matters most.
What is the CEO's part in all this?
You do not need to be the product owner. You do need to make the role real. That means three things: assigning it to a named person, protecting their time when other priorities push back, and holding business leaders, not IT, accountable for the results.
It also means getting close enough to AI yourself to judge the work. KPMG's survey found that 67% of leaders agree their CEO actively owns AI as a strategic business priority. In a smaller company, owning it is simpler. You can see every workflow, know every leader, and decide in one meeting what a large company decides in six.
If you are not sure who should own what in your company, Journey offers a free CEO AI audit to help you see where you stand and which one or two moves would matter most next.
Frequently Asked Questions
No. IT should own security, data access, integrations and vendor review, because those are what IT is measured on. The business should own the results, such as faster quotes or better margins. Between them, a named product owner should turn each business problem into a working AI workflow and drive adoption until people use it without reminders.
An AI product owner is the person who decides what an AI solution should do, for which team, and how you will know it worked. They map the current workflow, choose the lightest tool that fits, run short pilots with a stop date and track real use. In a smaller company this is usually an assigned role, not a new hire.
The business leader whose team uses the AI should be accountable for the result, with the metric written into their goals. The product owner is accountable for the solution and its adoption. IT is accountable for keeping it secure and connected. The CEO decides which problems get attention and holds each owner to their part.
Most companies of 25 to 500 people do not need a Chief AI Officer. They need a clear split of ownership and one person with protected time to act as AI product owner, often an operations leader or a respected manager who already uses AI. An outside partner can fill the gap if they leave real capability behind.
You can set it up in about a quarter. In the first week, write down who owns what. In the first two weeks, name the product owner. Then pick one or two problems tied to a number you report, run a pilot inside the real workflow, and review it around week twelve to scale, fix or stop.
