By Guest contributor |
Gartner forecasts that 40 per cent of enterprise applications will embed task-specific AI agents by 2026, up from less than five per cent in 2025. But speed of adoption is only half the story. The analyst also predicts that over 40 per cent of agentic AI projects are at risk of cancellation by 2027, citing inadequate risk controls as one of the key reasons why.
For the executives arriving at ServiceNow’s recent Knowledge 2026 event, the tension between the pressure to deploy and the need to control is the defining challenge of the moment. Ashlea Miller, senior manager of executive and employee communications and marketing at ServiceNow, sat down with Microsoft’s Silvio Bessa, who leads the global partnership between the two companies, to explore the evolution of the alliance and understand why the enterprises that master AI governance now will be the ones that define what a frontier firm actually looks like in the future.
The partnership between Microsoft and ServiceNow has been evolving for years. Can you tell us how?
ServiceNow is a particularly strategic partner for us at Microsoft. We believe in the strategy, we believe in where ServiceNow is going as a company – and Microsoft wants to be part of that. The partnership has moved from supporting IT service management as a system of actions, through to supporting AI, and now ServiceNow’s AI control tower (a centralised governance layer for managing AI agents across the enterprise) and the governance data around that.
The ServiceNow AI control tower has been a big topic of conversation at this event. What does it do for customers when you combine it with Microsoft Agent 365?
Every customer comes to AI with a different context, landscape and set of regulatory pressures. That results in a lot of AI chaos – customers deploying agents everywhere, from a simplistic business process through to an entire chain. This needs to be governed. We need some sort of AI governance to ensure security, AI robustness and loyalty to the data. When combined, Agent 365 and the AI control tower provide that end-to-end capability – the system of intelligence meeting the system of actions.
Microsoft also talks about ‘frontier firms’ – human-led organisations where agents are doing a lot of the work. What does that look like in practice?
Right now, organisations aren’t just managing people, they are managing people and agents. If we look back a year ago, we were talking about simple chatbots. Today, agents are primarily activating business processes and doing the work itself. Frontier enterprises are those that will engage with AI to transform their business model, to expedite and provide more customer satisfaction. Each company has a different journey to the frontier, but they all have a north star. And companies need to control what’s happening out there. A lot of IT professionals have one key question in mind: ‘How do I govern this – and how do I ensure security, performance and data integrity across all of it?’ I think ServiceNow plays an important role in governing companies on their journey to become a frontier.
Beyond the control tower message, what’s the bigger takeaway from the Knowledge 2026 event?
I had several meetings with executives during the week. Most of them aren’t just sitting in AI chaos – they’re in their own chaos. They have a lot of systems and want a better understanding on how they can integrate all of them.
When we talk about autonomous IT and the control tower, things are moving into the business process layer. I had a meeting with a customer about quality management – they want to integrate quality management of their product into IT, so that agents can operate across it. I even learned from a customer that they’re considering merging HR, IT and finance into one single entity. What I’m taking away from Knowledge 2026 is that IT is becoming more advanced to govern the entire AI estate – and this is not IT only. This is actually embracing the entire enterprise end-to-end.
A year from now, how will agentic AI have evolved?
For the past few years, the software has been dictating AI to the users. The software-as-a service components, the business process systems – they dictated how AI showed up. I think a year from now, we’ll probably see that pyramid reversed and the users will be driving the agentic behaviour into the systems. This means more personalisation of roles and of how people actually work – I may do the same job as my colleague, but we have different working patterns and agents will adapt to this. For example, Microsoft launched Frontier Tuning, a reinforcement learning environment that allow customers to get better results faster and cheaper by continuously adapting models and harnessing their data and workflows from real-world use. In other words, agents will drive personalisation and continuously learn based on the rubrics, patterns and skills associated with individual tasks or collections of tasks.
I also see a lot of the agentic processing moving into the presentation layer on the user’s device as GPUs become embedded. You can see the Microsoft Windows PC announcement with Nvidia RTX Spark recently as an example.
Discover more insights in interviews like this in the Summer 2026 issue of Technology Record. Don’t miss out – subscribe for free today and get future issues delivered straight to your inbox.