AI Ops… ready or not, it’s here!

All too often we get excited by the technology and try to jump straight into an AI Ops project without a solid plan or data check. This blog is a recap of a recent itSMF UK AI Ops webinar by all-round rockstar and fellow itSMF UK board member Val Wilson.

Why do we have to transform?

Val opened the ‘lunch and learn’ session by talking about the importance of taking the time to plan and organise so that our AI Ops implementations go well. The key message was that, before organisations implement AI successfully, they need to get ready for it.

She used the BT Group’s journey with AI Ops as the basis for her session. In the case of BT, they had a new CEO who was completely focused on transformation. The reality is that it’s 2026 and customers and markets are demanding AI and innovation; if it’s not the current offering the question becomes “why not?”. For BT, it became not just about a nice to have – it became about surviving and indeed growing in the current market.

The current landscape

BT has a large and complicated service ecosystem (or as Val put it, “it’s quite the beast”). To add to the pressure of the situation, many customers are critical national infrastructure customers who are the heartbeat of UK and therefore failure is NOT an option. This infrastructure is supported by over 2.5 thousand colleagues making up the service desk, technical teams and all the wider ITIL teams. BT have been migrating to ServiceNow for the last 18 months and are in the latter stages of what has been a long migration to this strategic tool.

The organisations that succeed with AI won’t be the ones that move fastest. They’ll be the ones that prepare properly.

Val continued by talking about how having a solid business case will set you up for everything you know and love when you go into delivery mode. BT partnered with Accenture to ensure their AI Ops journey had the best possible start, and one of the key areas of focus was getting the vision right. As Val put it, what vision do you want to achieve, what problems do we want to solve, and how will we get there? The business case should be a living, breathing document with referenceable documentation that will help people go back and review.

Understand patterns, performance and pain points using trusted data.

Before introducing AI Ops, it’s essential to understand the reality of your current service ecosystem. This means building a picture from trusted data rather than assumptions. Look at the fundamentals: your addressable support base, staff locations, ticket volumes, technology landscape, recurring incidents, closure codes, alert volumes, MTTR, MTBF, customer feedback, user sentiment, running costs, architecture, and even future revenue forecasts. The goal is not simply to collect data, but to understand the patterns, performance trends, and operational pain points that exist today.

Even in highly mature environments like BT, there is often more complexity than we expect. In Val’s case, BT were not starting from scratch as there was extensive automation and zero touch ticketing, yet there was still work needed to understand how services worked at a component level, service mapping and what operational friction existed, and what incidents were generating the biggest pain points.

Start with high impact, achievable use cases that deliver value.

Now we’re going to use those data insights to focus our use cases. In the case of BT, knowledge came from lots of different places, for example, service management tools, monitoring platforms, documentation, support teams, and business stakeholders. A key priority is to understand your workflows, broken down by tasks so you know exactly how they work before you try to automate them. Working with ServiceNow, BT employed a sandbox environment with different use cases to demonstrate proof of value for one customer, and used practical guidance from ITIL 5 and visualised how AI would work with the SXP platform.

Engage the right people, align priorities across the organisation.

With AI Ops there are lots of different stakeholders and we need to look after them all:

• Strategic;  there are many doors to get through and the right stakeholders will help you open them
• Operational; the colleagues that will help you figure out what works in real life
• Technical and delivery; the people who do the day-to-day work so it’s really important to keep them close.

It’s important to set expectations early on. AI isn’t a quick win ‘one and done’ exercise. It’s typically a 2 to 3 year programme so when we introduce AI Ops we need to get some quick wins so that we can deliver some value quickly. Some prerequisites for success are:

• Trusted, accurate, connected data for mature processes is essential for AI to work
• Service mapping so you can understand events and how they correlate
• Incident and problem resolution information.  Do a refresh of your knowledge and check to make sure it is up to date and ready for AI.
• Measure value, outcomes and operational impact continuously
• Look at getting the revenue right – will implementing AI Ops get you more business?
• Get feedback from customers and insights such as NPS to help drive you forward
• Pragmatism is needed so we can demonstrate value – let’s just move forward and finesse it as we go
• Do your programatics – what are the big tech things that you do that give you the best outcomes? We need to do this to grow and be competitive.

Build a culture that embraces change

Adoption and early life support can wreck your digital implementation because, if people think they won’t have a job, then adoption will stop so we need to have the important conversations first. To this end, BT runs  AI apprenticeships which are  supported by the government, helping colleagues who could be displaced by AI move into different roles. This can give people different pathways in a secure environment.

How can we derive more value from AI Ops?

Here are Val’s top tips:

• From greenfield to RFP, it’s 6 to 9 months
• Do the prep work, for example making sure your data is solid and you have the right people engaged
• Make your service catalog modular
• Train your sales guys to talk about AI Ops
• Look at where we have adjacencies and where we can deploy AI in the wider business.

That was my take on Val’s ‘lunch and learn’. What do you think? Please let me know in the comments.