It is a common pattern. A company decides to implement AI. They assign the project to IT and expect IT to make it happen. IT evaluates tools, selects a vendor, sets up the infrastructure, and gets the system running. Then they hand it over and expect the business to use it.

Often, nothing happens. The system is installed but not used. Or it is used inconsistently. Or people use it but get poor results because they are using it wrong. The AI implementation fails, and the company blames the technology.

The problem is not IT. IT did what IT does well: they delivered a functioning system. The problem is that AI implementation is not primarily a technology project. It is a business transformation project that happens to involve technology. And business transformation requires more than IT can provide alone.

The Limits of IT

IT is responsible for several things. Selecting the right tool based on technical requirements. Installing it correctly. Ensuring it integrates with existing systems. Keeping it secure. Maintaining it over time. These are necessary, but they are not sufficient.

What IT cannot do is define the business problem being solved. IT can tell you whether a system is technically capable of solving a problem, but they cannot tell you whether the problem is the right one to solve. They cannot tell you what the business impact will be. They cannot convince people to change how they work.

IT can implement a system. They cannot drive adoption. They cannot change people's minds. They cannot manage the organisational change required to use the system effectively. They can train people on how to use the system, but they cannot make people actually use it or use it well.

This is not a criticism of IT. IT is doing their job. The problem is expecting IT to do jobs that are outside their scope and expertise.

Business Problem Definition: Why This Problem Matters

The first critical input to AI implementation is defining the right business problem to solve. This cannot come from IT. It has to come from operations and business leadership who understand what actually matters for the business.

An operations manager knows which processes are most painful, which ones have the biggest impact on customer satisfaction or profitability, and which ones are ready for automation. An IT person can see that a process is complex, but that does not mean it is the right one to automate with AI.

Worse, if IT chooses the problem to solve, they will often choose based on technical criteria rather than business impact. They might choose something that is technically interesting but not particularly valuable to solve. Or they might choose a problem that could be solved more simply with a different approach.

Business leadership needs to define what problems matter. Then IT works on technical solutions to those problems.

Process Understanding: How It Actually Works Now

Before you automate something, you need to understand how it actually works. Not how the documentation says it works, but how people really do it. What steps do they take? What decisions do they make? What data do they use? What exceptions occur?

Operations and process experts understand this. They have watched the process happen hundreds of times. They know where people get confused. They know what information is missing. They know what usually goes wrong. This understanding is critical for designing an effective AI system.

IT does not have this understanding. They can look at the documented process and the current system, but they do not know how people actually work. When they try to design an AI solution based on incomplete understanding of the process, the result often does not work well because it does not match how people actually need to work.

Change Management: Getting People to Adopt the New Way

The biggest barrier to AI adoption is usually not the technology. It is people. People are comfortable with the current way of working, even if it is painful. They do not want to learn a new system. They do not trust that it will work. They worry about what the change means for their job.

Managing this change requires communication, leadership, and patience. People need to understand why the change is happening. They need to see leadership committed to it. They need training that makes them confident using the new system. They need support when something goes wrong. They need to see success from early users and know that the change is worth the effort.

This is fundamentally a people and organisational challenge, not a technical one. IT is not equipped for this. Organisational leaders and middle managers are. They are the ones who can model the new way of working, answer questions about why the change is happening, and support people through the transition.

Expected Business Impact: What Success Looks Like

Before you implement an AI system, you need to define what success looks like. How much time should be saved? What should the quality improvement be? What should the cost reduction be? What should happen to customer satisfaction or employee satisfaction?

These definitions need to come from operations and leadership, not IT. IT can tell you what metrics the system can capture, but they cannot tell you what targets are realistic or what business outcomes matter.

If you do not define success upfront, you end up evaluating the AI system against unstated expectations. One person thinks it should save ten hours per week. Another thinks it should eliminate errors completely. Another thinks it should be completely hands-off. When it does not meet all these unstated expectations, it is considered a failure.

Clear, measurable success criteria, defined upfront by operations and leadership, prevent this problem.

The Right Structure: Cross-Functional Team

Effective AI implementation requires a cross-functional team. IT handles the technical layer. Operations and end users provide input on the current process, the business problem, and what success looks like. Leadership provides guidance, sets direction, and ensures adoption is prioritised.

A strong AI implementation has a project leader who is not from IT. This person, typically from operations or management, owns the business outcome. They define the problem. They ensure the AI system is designed to match how people actually work. They manage the change and drive adoption. IT supports this leadership by implementing the technical solution.

When implementation is structured this way, with clear roles and cross-functional input, success rates are dramatically higher. When IT is left alone, success rates are lower because key business input is missing.

The Reality: You Need More Than One Team

This does not mean IT is not important. IT is essential. You cannot implement AI without people who understand technology, can select the right tools, can ensure systems integrate correctly, and can maintain everything over time. But IT alone is not sufficient.

You also need operations people who understand the current process and can identify what needs to change. You need leaders who can prioritise the right problems and drive change through the organisation. You need end users involved early so the system is designed to match how they actually work, not how IT imagines they work.

The companies with the best AI implementation success are the ones that assemble a team with representation from all these areas. IT, operations, leadership, and end users all playing their part. That is what makes AI implementation succeed. IT alone, no matter how good they are, cannot do it.