In the AI gold rush, I see two distinct categories of solutions: vitamins and painkillers. A vitamin is something that helps you write better copy or polish a slide deck, which only boosts your productivity so much. A painkiller, however, is a matter of systemic survival. It can help you fix an issue before it gets out of hand or deal with a global outage, like when one CrowdStrike change crashed roughly 8.5 million Windows PCs in July 2024.
When embracing AI, it’s better to go after the painkiller category than the vitamin one. I’m not surprised that Gartner predicted that in 2026, AI will evolve “from tools that assist humans to platforms that replace manual effort for complex workflows.” Imagine how painful that CrowdStrike issue was and how much better AI agents could have handled it. IT automation is a painkiller.
IT is akin to firefighting
The painful truth is that the IT industry is drowning in data but starving for insights. There’s enough data for you to collect, but is that data giving you answers? This “data noise” has always been the issue in IT, turning it into a perpetual state of firefighting.
The rest of the business doesn’t care about IT operations. Sure, IT provides helpful dashboards, but those are more like insurance: they offer peace of mind that the business is protected. Meanwhile, IT employees are the first responders trying to put out fires that can spread quickly. There are plenty of high stakes, millions of dollars on the line, and they can’t always deal with it all. It’s incredibly hard work.
If a fire gets out of hand, can AI tools douse it and make sure that it doesn’t lead to a bigger catastrophe? In my experience, the answer is a resounding yes. The business cares about IT fixing the problem, not about speeds, feeds, and what’s under the hood. IT is a value-based, outcome-based use case for AI.
Give agents more agency
Put another way, AI makes self-healing possible in IT, shifting the work from firefighting to fire prevention. You don’t want humans spending time firefighting, like they did 20 years ago. Today, you want smoke detectors in your IT infrastructure. When there’s smoke, the AI agents detect it and go fix it. There’s no need to call the fire department and deploy firefighters every single time.
Agents can help you cut down noise and investigate issues. For example, “Bob complained his app is running slowly. No problem, have the AI agent investigate and come back with the root cause.” A second agent fixes the issue, and a third agent builds a dashboard for Bob that shows his team what’s really going on. It’s no wonder that Forrester believes multi-agent systems are the next advancement in agentic AI. You can have a swarm of agents working together to detect, investigate, resolve, and learn. That’s the whole loop. Over time, as AI takes over, the percentage of mundane IT work shrinks, and humans become reviewers.
Human workflows can be complex, like a customer who booked a flight, experienced an issue, and wants a refund. These are hard things for agents to handle. Tasks like reading logs? Agents do that in their sleep. In fact, they never sleep. IT automation is an apt use case for agents.
Once you shift your mindset, you’ll want to give your agents more agency by ensuring they understand all your incidents, services, locations, and the org structure. Ultimately, the more your agents understand your company’s data relationships, the more they can accomplish. It’s not just the what. It’s the why. “Why did Bob fix this router flapping issue a certain way? Well, here’s what he did.”
There’s structured knowledge, messages in Slack, and calls in Teams. You can bring all that together in a context graph and feed it into your agent engine that mimics what humans do: sense, plan their work, reason over it, take action, and adapt.
Two years ago, chatbots just spit out an answer. It was either really good or really bad. Since then, we’ve added reasoning and action. That’s where the whole loop comes to life. Giving agents agency is a new way of building, managing, and fixing IT applications.
Measure how agents are helping
Across all our customers who want IT automation, the top no-brainer use case is noise reduction. There’s so much noise happening, and nobody likes getting woken up at ungodly hours, especially when most of the time there’s no fire, just smoke.
You don’t want to waste human labor on smoke alarms. You want the agent to handle that. Noise reduction is a clear value KPI. If you reduce noise by 90%, fewer incidents enter your IT Service Management (ITSM) tool, like ServiceNow. Fewer tickets in ServiceNow means your IT team has more time to focus on critical work. That’s the top metric our customers measure: fewer incidents entering their ITSM.
The second critical metric is time to investigate. Investigations used to take four to eight hours on average. “Hey, Bob, go investigate this issue.” Every time you ask Bob to go investigate, it takes Bob time. You can’t ask Bob to investigate every sign of smoke because it will take too long. Bob comes back with a report after eight hours. As a result, if you don’t have agents handling most of your investigations, you have to let some fires burn. The cost is some fires become firestorms. You are flying blind. You see these pesky alerts coming in, and you have to let them be until you know the smoke is a fire. With agents, you can investigate every anomaly in minutes, and you can do infinite investigations.
Third, some customers track Mean Time to Repair (MTTR), which measures how long it takes to resolve an issue. They measure how many incidents are fully autonomously resolved. If the system generates an issue, it’s a system fault, and the agent can determine the issue, fix it, and tell you it fixed it. It’s all tracked in the ticket, so you can see the audit log. It’s as if a human did it, but agents are doing everything. MTTR nowadays can be around 10% to 20%, but that number will only grow.
Agents change how your company operates
Noise reduction, time to investigate, and MTTR are all reasons why we see autonomous IT resonating with our customers. Many don’t really have a Level 1 team anymore because they have completely automated that function. There’s just Level 2 and Level 3. In April, McKinsey estimated that agentic AI can eventually automate 60% to 80% of routine infrastructure work and noted an example in which 50% of a company’s service agent capacity was redeployed to higher-value activities.
After companies deploy the right AI tools, their IT teams transform how they work. That’s why IT automation is a painkiller: it’s a forcing catalyst for embracing AI agents across your organization and letting your human workforce keep your business running smoothly.
