AI is the Exam Your Organisation Hasn't Revised For

This article follows on from part one of our blog series, "What Did Your Shared Services Actually Change?"

Global Business Services (GBS) brings functions such as HR, finance, procurement and IT together to deliver services across the business. The aim is to support growth, control costs, serve customers while managing risk. Achieving that means agreeing how work passes between functions, who makes decisions and when the work is actually finished. AI puts those arrangements to a practical test. Through the example of opening a new office, Ashley and I explore what happens when AI reaches a hand-off that still depends on somebody knowing who to call.

Keyvan: Surely this gets us closer to opening that office.

We had made it to Las Vegas and through the first day of ServiceNow Knowledge. Ashley had taken his badge off. Mine was still hanging round my neck, backwards. We had found a table in the hotel bar where we could hear each other without leaning across it.

On the drive, we had been talking about a business planning to open offices in several regions. It had set up GBS to support that growth without its support costs rising at the same rate. I had spent the afternoon watching demonstrations that seemed to offer a way of doing that.

Keyvan: Procurement gets the equipment ordered faster. HR processes the starters faster. IT gets the accounts set up. There must be a fair amount of time we could take out of the whole thing.

Ashley: There could be. Now suppose the opening date changes. HR updates its plan, procurement is already working from the old one, and IT is waiting for an approval nobody realised it needed. Each function can get quicker at its own work. Someone still has to make sure they are working towards the same result.

Keyvan: We could make every department quicker and still open the office late. Which would be a difficult result to explain after all those demonstrations.

I had been adding up the time each department might save. The business was waiting for an office it could use, with people ready to serve customers. That was what we had wanted the GBS programme to make easier in the first place.

Ashley: That is what you need to test. You can have useful AI in each function and still need somebody chasing the work between them. If opening every office needs the same amount of that person's time, you have left part of your growth problem untouched.

Keyvan: But we have put someone in charge of shared services. Surely these are exactly the things they should have sorted out. Why does AI change the conversation?

Ashley: Because now you are asking a system to act on the agreements. When a person handles the request, they can ring someone, check what was meant and get permission to carry on. Ask AI to do that work and you have to explain which information it can trust, what it is allowed to do and who decides when the answer is unclear.

The work gets done, and the extra effort disappears into their day. I had been thinking about the tasks AI could take away from them. Ashley was asking what knowledge and authority needed to go into the service first.

Ashley: AI is an exam most organisations haven't revised for. It asks them to be clear about things they have managed to leave unclear for years.

Keyvan: Surely it can help with some of that preparation. Give it the records and let it see how people actually get things done.

Ashley: Yes. It can help find the patterns and the missing information. The people responsible still need to agree which way of working they want to keep. Something appearing repeatedly in the records might be a workaround everyone has learned to live with.

The barman brought two drinks with rather more fruit than I had expected. Mine had a small umbrella. Ashley's did not. I decided against asking why and took my badge off.

Keyvan: All right. What would you ask the person running GBS to show you before putting AI into those services?

Ashley: Four things.

He took a sip. I was still negotiating with the umbrella.

Keyvan: Four? That many?

Ashley: First, data the functions can trust and use consistently. Second, a process they have written down and agreed, from the initial request through to the result. Third, ownership, so people know who can decide what across those hand-offs. Fourth, a standard people follow, including when something is urgent or needs an exception.

Keyvan: Data I can picture. HR, IT and procurement need to agree which employee record to use and what the dates mean. The process ought to be straightforward enough. HR completes its tasks, procurement buys the equipment and IT delivers it. I would probably have called that finished.

Ashley: Can the person do their job? That is the test. A laptop delivered without the access they need leaves the business waiting. Your process has to reach the result the service promised, including what happens when a date changes or an approval doesn't arrive.

Each team could have evidence that it had done what was asked. The manager would have evidence too. A new colleague sitting beside them, waiting. It was a fairly practical test of whether GBS had made those functions work together.

Keyvan: Ownership sounds harder. Calling someone the service owner won't give them authority over HR's decisions.

Ashley: Exactly. Ownership needs an agreement behind it. The GBS lead and the functional leaders have to settle who can decide what, including the exceptions. The CIO can connect the systems. They cannot make HR's employment decisions or set procurement's financial approval limits.

Keyvan: And the standard has to hold when people are under pressure. I can already hear someone saying this particular office is urgent and asking a colleague to sort it out.

Ashley: That is the standard being tested. Required information gets checked. Approvals happen where they are needed. Urgent work has a route people can use, and exceptions go to someone authorised to decide. Otherwise you end up automating the official process while people keep the business moving somewhere else.

I could picture the workshop where everyone agreed. I could also picture the Monday afterwards, when a regional director rang someone they knew. The standard would have to survive that call.

Keyvan: The regions can have good reasons to work differently. Are we now telling the Japanese office it has to copy the UK?

Ashley: They can have different suppliers, equipment and local requirements. Agree what the rest of the service needs to rely on. What does an approval mean? What confirms someone is ready? Which information passes to the next team? Record the variations so the service can accommodate them. Otherwise you are asking somebody to interpret every region again.

Gartner makes a related argument in its March 2026 paper, CIO Top Practices to Redesign Shared Services for the AI Era. It argues that the organisation needs to own decisions about shared data, standards and platforms, with room for local adaptation. It warns that separate AI initiatives can fragment the very services they are supposed to improve.

A separate 2025 Gartner survey of 360 people involved in rolling out generative AI found only 14% were very confident their content and data were suitably secured and governed to provide value for AI and human use. The finding reflects those respondents' confidence across enterprise applications.

Keyvan: So the GBS programme needs to make those shared decisions hold across the regions. And the enterprise service management work is how we turn them into something people can actually use.

Ashley: Yes. Define the service, connect the work and build the agreed rules into it. The platform carries information between teams and shows what is waiting. Then decide where AI can handle a request or take an action, and where it needs to stop. Getting the systems connected and the data right is real work too.

That gave the CIO something specific to build with the people running shared services. There was still a substantial technology job. There were also decisions the GBS programme had to bring into the same room.

Keyvan: I would want to see what happened afterwards as well. If someone has access they shouldn't have, I need to know what information was used, who authorised it and what the system did.

Ashley: Build that record into the service from the beginning. Keep the inputs, permissions, actions and human approvals visible. Being ready includes knowing when judgement is needed and who takes over. That is how you give AI a defined job and keep responsibility for the result.

Keyvan: This could still turn into another year of redesign before anybody gets anything useful. The business wants that office open.

Ashley: Start with a manageable service within that plan, such as getting the first team equipped and ready to work. Bring the functions involved together around it. Work through the data, process, ownership and standard for that scope. Test the awkward cases. You can begin with a small amount of automation and a clear place for people to take over.

I opened my notebook. The first question I had written before we reached Las Vegas was which version of the business we were going to give AI. I added another: could the next office use what we were about to build?

Keyvan: And we judge it by what changes for the business. Can we support the growth? What did it cost? Can we keep the promise to the customer? Have we kept the necessary controls?

Ashley: Yes. Measure the whole service, including the waiting and chasing. Then see what you can reuse for the next office. A reliable employee record, agreed approvals and a clear definition of ready give you something to build on. That is progress towards the GBS promise of supporting more business without repeating all the effort.

I could see a first project worth doing. I could also see a question that the person running GBS would have to take back to several colleagues.

Keyvan: HR needs to do some preparation. IT needs to connect things. Procurement has work to do as well. The benefit reaches the new office and its customers. Even with everyone agreeing on the service, I can see each budget holder asking why the preparation should come out of theirs.

Ashley: That is where these conversations get stuck.

The barman came to collect the empty glasses. We had talked about walking part of the Formula 1 circuit before we left Las Vegas. I wanted to see whether I could recognise any of the corners at walking speed.

I closed the notebook. We still had to work out who would pay.

Keyvan Shirnia is Chief Revenue Officer at Fusion GBS.

Ashley Gionnotti is the ServiceNow Technical Delivery Lead at Fusion GBS, based in London.

This blog is part of The Six Critical Capabilities series.

To find out where your service management foundations actually stand, the Fusion GBS scorecard gives you a benchmarked baseline in five working days. It is the front door to Value Adoption Services: use the data to find low-maturity quick wins, link the first improvement to a business outcome, secure modest funding, implement it, measure and benchmark the value, then use the evidence to support the next step.