The evolving role of AI in ServiceOps

Haroon Ahmed, Vice President, Customer Success and Pre-Sales at BMC Helix, discusses the company’s sharper focus on service and operations management, how agentic AI is transforming enterprise workflows, and why human judgement, interoperability and governance remain essential as organisations scale AI adoption.

How has the transition to BMC Helix changed the company’s product strategy and go-to-market approach?

If you think about the traditional BMC business, it had mainframe, enterprise automation with Control-M, and then the division I work in, which is service management and operations management. That has now become BMC Helix. They’re different businesses serving different audiences. When you look at where AI is most applicable from a growth perspective, it’s really in our space, in service management and operations management. Spinning off the company gives us several advantages. When you’re part of a large organisation, the focus can get diverted. Now we have a clear focus, and we have a CEO, Ali Siddiqui, who comes from an engineering background and brings a startup mindset to the company. We’re not really a startup, of course, because we already have an established customer base, but we’re now very focused on ServiceOps and applying AI technology to this space.

About five or six years ago, we made a fundamental decision to change our architecture and move towards containers. That allowed us to innovate faster, move faster and build technology more quickly. We also started embedding AI into our products. We had conversational and generative AI use cases within our solutions even before ChatGPT became widely known.

How is AI integrated into the everyday workflows of service management and operations teams? Is it primarily a conversational interface, or does it go beyond that?

AI comes in different forms. One is the conversational experience. For example, you can use Microsoft Teams to have a service management or operations-related conversation with our solutions. We’re not asking customers to use a separate collaboration tool. They can continue using Teams and interact with ServiceOps capabilities through that interface.

But AI is also embedded directly into everyday workflows. Take major incident management. A service desk agent working on a ticket needs to understand its context, including the work logs, conversations between support teams and users, and previous attempts to resolve the issue. Traditionally, the agent would have to go through all that information manually. Now AI can automatically summarise the context, look at previous tickets, identify earlier recommendations, perform sentiment analysis and suggest how to resolve the problem. So while there’s a conversational component, AI is also integrated into the workflows that knowledge workers use every day, helping them resolve issues faster and more efficiently.

With AI automating many of these processes, what level of expertise will human teams still need, particularly when it comes to making decisions?

You still need human expertise. When you think about AI solving a problem, there’s the concept of a human-in-the-loop, where the person is actively involved in performing tasks and connecting the dots. Then there’s human-on-the-loop, where AI does much of the work, such as summarising information or analysing sentiment, but leaves the judgement to the person. That’s an important distinction. If you look at the evolution of the knowledge worker, we’ve moved from physical work to knowledge-based work, where people use tools to perform tasks. AI takes that further by removing much of the repetitive work. If somebody previously had to go through 50 clicks in an application to reach a conclusion, AI can now analyse the relevant data and perform those steps because the process is predictable.

But once AI reaches a conclusion or develops a hypothesis, somebody may still need to make a judgement. If AI has made the right decision the last 100 times, perhaps you’ll trust it enough to automate that decision. But where the confidence level isn’t high enough or the margin for error is unacceptable, you’ll still need a human-on-the-loop. That’s where human judgement and intuition come in. I often use an analogy from my experience working at a bottling facility. The machines did the actual bottling, but I was there to operate them, make sure everything was running properly and intervene when something went wrong. The same principle applies to AI.

But if employees are no longer directly involved in many of these processes, could their expertise and awareness diminish over time, affecting their ability to make the right judgement when required?

I think they’ll be able to make those judgements, and I’ll give you an example. I run the pre-sales team, and I’ve been hiring quite a few new graduates because they’re AI-native thinkers. When we ask them to learn a product and prepare a demonstration, it’s amazing how quickly they can deliver a good demo. AI helps them learn much faster. It’s almost like having access to five years of experience that they can tap into immediately. They have access to vast amounts of knowledge, and AI helps them connect the dots. But they still need to know how to ask the right questions and provide the right context. That’s a skill, and I believe AI-native people are developing it naturally.

There may be gaps in their knowledge, but the information is at their fingertips. As long as they know how to access that knowledge and guide AI towards the right output, they can become more effective. I see this in my own work. I’ve never considered myself a particularly good writer, although I’ve written blogs over the years. It used to take me a long time. Now I can put my ideas into bullet points and ask AI to develop them into an article. But I still guide it, tell it what to include, what to avoid and how to reflect my writing style. The ideas remain mine. AI simply helps me express them more effectively. I think the same principle applies to other areas of work. You may not have all the expertise yourself, but if you know how to use AI to get the right information, you can make better decisions.

What kind of feedback have you received from customers in the region over the past few months, particularly around your AI capabilities?

I’ll give you an example from the event. One of my colleagues was sitting next to a senior director from a government agency here. After seeing the demonstrations, he wanted all those capabilities implemented in his organisation. That’s the kind of reaction we’re seeing because these are real problems customers are trying to solve. They want to improve employee experience and reduce the workload on their help desks. Every time an employee has a poor experience, they create a ticket, and that means more work for the service desk. Customers recognise that AI can help address these challenges. In some cases, they’re also working towards automation targets set by their leadership or government authorities, so there’s a strong motivation to adopt these technologies.

At the same time, we’re being very deliberate about the agents we develop. We’re not trying to create a thousand agents or build a sea of agents. We’re focused on ServiceOps, and our agents are designed to solve specific problems in that space. These include employee experience, major incident handling, recurring issues, root-cause analysis and change management. We want to help organisations manage changes without creating downstream problems and improve the quality of their operations. Customers are receptive because these agents address problems they encounter every day.

How will BMC Helix solutions coexist with other enterprise AI ecosystems, including organisations developing their own agents or using AI capabilities from different vendors?

The key is adherence to industry standards such as Model Context Protocol (MCP) and Agent2Agent (A2A). We support these standards today, and that’s how we’re going to coexist with other systems. Whether we’re accessing another system or another system is accessing ours, these protocols enable that interaction. It’s not going to be a monopoly, and it cannot be a monopoly. We focus on IT ServiceOps, but we also need to work with the other systems that customers use.

I’ll give you an example from my own work in customer success. I use Gainsight for adoption-related activities and Salesforce for sales opportunities and other information. Through MCP, systems can expose their capabilities and connect with different AI interfaces. I could use Claude, ChatGPT, Microsoft Teams or another interface to access the information I need. These standards make it much easier to connect systems and bring together information at the right time and with the right context to support decisions. I’m genuinely impressed by how powerful this way of working has become.

AI investment is still evolving, but are enterprises already moving beyond experimentation? What challenges are they facing as they scale their deployments?

I don’t think we’re at an infancy stage anymore. I go to customers who tell me they already have 2,000 agents. The issue isn’t that organisations aren’t building agents. They’re building plenty of them. The challenge now is that different teams are creating agents to improve their own productivity, but when organisations try to bring everything together, they realise there’s no common governance, security or guardrails.

That’s what customers are dealing with now. How do you bring some order to all of this? How do you establish governance, security and guardrails, and ensure that agents work with secure systems of record? AI can hallucinate, and that creates risks when systems are allowed to act without sufficient controls. So I think the next challenge is not simply deploying more agents. It’s managing and governing the agents that organisations already have.

Does BMC Helix provide its own orchestration and governance framework, and can customers use it to develop custom agents?

Yes. We have Helix Agentic Engineering, which allows us to build and run these capabilities, and customers can also develop their own custom agents.

We currently have around 30 agents. I can’t say exactly how many we’ll eventually have, but it’s not going to be thousands. As I mentioned, our focus is ServiceOps, and we’ll continue developing agents that address the critical problems within service and operations management.

Can you give some examples of the different types of agents and the specific tasks they perform within ServiceOps?

I would describe them in terms of the problems they’re designed to solve. Take employee experience. An employee might want to know the organisation’s maternity policy or vacation policy. Traditionally, they’d have to search the intranet or contact HR. We have an agent called Employee Navigator that allows them to ask the question directly and get the information. Or suppose an employee wants a new laptop. Instead of searching for the right form and submitting a request manually, they can simply say they need a laptop. The agent can initiate the service request, ask whether they want Windows or Mac, determine the required specifications and even check whether they’re eligible for an upgrade. That reduces the number of tickets reaching the service desk. When tickets do come in, AI can identify those relating to the same underlying problem and group them together rather than having agents work through a hundred similar tickets individually. It can also summarise tickets, analyse sentiment, categorise issues and route them to the appropriate teams. All of this reduces service desk fatigue.

Then there’s the operations side. You have teams managing networks, applications, infrastructure and cloud environments. If a problem occurs, AIOps can detect it, identify the affected service and determine the likely root cause. For example, a mobile banking application might be experiencing problems because of an underlying network issue. AI can identify that relationship without requiring several different administrative teams to investigate independently. It can then associate the root cause with the relevant service desk tickets, retrieve knowledge articles describing how to resolve the problem and, where appropriate, trigger an automation, such as restarting a network interface. The service desk and operations teams are interconnected through these workflows. We’re also measuring the outcomes. When agents perform work, the system reports how many hours have been saved and the associated cost savings. That gives organisations visibility into the capacity they’re recovering and allows employees to focus on other work.

There are several established players in IT operations and monitoring, many of which are introducing AI capabilities. How does BMC Helix differentiate its approach?

Our focus is really on the enterprise and solving complex problems. One of our biggest differentiators is openness. We’ve always taken an open approach because we recognise that customers operate heterogeneous environments. If a customer uses ServiceNow for ticketing, that’s fine. Our AIOps agents can still work with that environment. If they’re using Dynatrace for application monitoring, SolarWinds for network monitoring or Microsoft’s SCOM, that’s not a problem either. We don’t want to be excluded simply because a customer isn’t using our ticketing or monitoring products. We can bring information from these different systems together to solve the operational problem. Ultimately, we’re trying to do three things: identify the root cause quickly, recommend how to fix the problem, and help execute the automation. That’s what reduces mean time to resolution and improves the end-user experience. Even if a customer isn’t using our ITSM or monitoring solutions, we can still help them with AIOps. That’s an important differentiator for us.

The second is our ability to operate at enterprise scale. There are relatively few companies operating at that level with the trust of large enterprise customers. The third is our focus on ServiceOps. We’re not distracted by trying to do everything. We’re focused on bringing service management and operations management together. That’s where we believe our strength lies, and it’s also an area where industry analysts have recognised our capabilities. When you combine that focus with our open architecture and the pace of innovation we’ve maintained, particularly around AI, I think it puts us in a very differentiated position.