Where AI is actually working in IT workflows today

Need content here!

In production

The hype around AI is loud. But underneath it, something quieter is happening — in some organizations, in specific workflows, AI is already doing real work. We asked 13 IT practitioners where they are actually seeing it land. Not in the roadmap. Not in the pitch deck. On the floor, today. Here's what they told us.

Incident correlation Status: proven, where the data is there

Before — Five incidents fire within ten minutes. Five different agents pick them up, each working their own ticket, each using slightly different language to describe what they're seeing. Nobody connects the dots. Somewhere beneath the noise, a major incident is forming.

After — Where AI is in place, it can read across multiple incidents simultaneously, regardless of how each one was described, and flag patterns that human teams might not catch in time. The major incident gets raised earlier. The team has a better chance of getting ahead of it.

"AI tools have the ability to look through this, even if they're using different language, different words to describe what's happening, and say — hey, we might have a problem here." — Ken Gonzalez

Auto-resolution of known events Status: proven, for well-defined failure patterns

Before — A database fills up overnight. No alert gets actioned until morning. By then it's an outage, a ticket, a war room, and a postmortem.

After — For failure patterns that are well understood and well documented, AI can detect the issue and resolve it without a technician having to manually step in.

"AI monitoring tools are identifying events and closing incidents before a human sees them. Think of a database filling up and AI takes care of that, resolves that incident without anybody touching it." — Jeffrey Tefertiller

Ticket routing and classification Status: proven, though accuracy depends on data quality

Before — A ticket comes in, gets miscategorized, sits in the wrong queue, gets rerouted, and eventually reaches the right person two days later. The user has already called three times.

After — In organizations where this is working well, natural language processing reads the description on arrival, classifies it, and routes it to the right team without manual intervention.

"Automated ticket routing and classification based on the content of the description — I could not only prioritize it properly, I could get it to the right people to get that incident tended to more quickly." — Phyllis Drucker

"Automatically triaging incoming tickets — that's a big one where AI can help." — Stephen Mann

Self-service and knowledge management Status: proven, with wide variation in how far organizations have taken it

Before — A user has a question. They log a ticket. They wait. A level one agent reads from a knowledge article that was last updated eighteen months ago. Everyone's time gets spent on something that should have taken two minutes.

After — Where chatbots and virtual agents have been properly implemented, users can get answers, guidance, and in some cases automated fixes without ever needing to log a ticket.

"Many organizations are dabbling in that area and finding that it provides tremendous value in handling the repetitive questions level one technicians get and getting people back on their feet." — Phyllis Drucker

"Where a user is able to log in and get meaningful assistance with the issue they're facing — that's where we're seeing a lot of value." — David Cannon

"Many users will do anything to avoid clunky conversations with service desk agents who seem to be doing their best to impersonate robots." — Mark Smalley

Automated summaries and communications Status: proven, mostly in larger or more mature IT operations

Before — An IT professional picks up a complex ticket mid-shift. They have to read through the entire history to understand what's happening. When the user calls for an update, someone has to manually draft a response.

After — In operations where this is running, AI can surface a summary of the ticket history before the agent opens it, and generate follow-up communications to keep users updated without someone having to write them.

"AI has been helping by producing automated summaries that help IT professionals gain better situational awareness around the things that customers need them to help with. Second thing is producing good follow-up communications to help those that they serve stay updated on where a resolution is at for them." — Ken Gonzalez

Patching Status: proven, though adoption is uneven

Before — Patching gets done differently by different teams, on different schedules, with different standards. Inconsistency creates gaps. Gaps create risk.

After — Where AI-assisted patching has been adopted, it can bring more consistency to a process that has historically resisted it, reducing some of the gaps that lead to vulnerabilities and outages.

"Patching is a great one where everybody wants to do it a different way. Those kinds of things — getting more consistency — gets us to a more reliable infrastructure, which gets us to fewer interruptions." — John Custy

Monitoring and observability Status: proven, and arguably the most mature use case

Before — Humans can only watch so many dashboards. Alerts fire constantly. Critical signals get lost in the noise. Someone eventually notices something is wrong because a user calls to say it is.

After — AI can monitor across large infrastructure environments in real time, correlate signals from multiple systems, and surface the alerts that actually need a human — reducing the noise that consumes so much of operations teams' time.

"Security is absolutely becoming far more autonomous. We're seeing it in terms of scalability and being able to scale up and scale down resources. We're seeing it in terms of auditability and being able to monitor in real time what is happening in broad systems." — John Santaferraro

"The ability to quickly spot patterns, guide your agents, your operations agents towards what's actually happening, separating signal from noise amongst all that data — that is a great use case for AI today." — Shane Carlson

"Monitoring that is able to look at different things, identify correlations, and alert the right person rather than just a team." — John Custy

Documentation and knowledge navigation Status: emerging, but showing early promise

Before — A hundred applications, two hundred pages of documentation each. Twenty thousand pages. An ops team member needs to know which process applies to which system, right now, under pressure.

After — AI can help navigate that scale of documentation in ways that weren't practical before, surfacing relevant information faster than manual search.

"Where I really see success and value is in helping operations staff manage documentation and figure out where they need to go in order to do the thing they're intending to do. AI can really help us in that space." — Ryan Schmierer

Security operations Status: emerging fast, and the stakes are high

Before — A zero-day exploit was discovered. It typically took weeks before it was seen being actively used against infrastructure in the wild. Security teams had time to respond, plan, patch.

After — That window has collapsed. In some environments, AI is helping security operations teams keep pace — watching attack frameworks, processing threat data, and helping analysts understand what they're seeing faster than they could manually.

"Something that used to take weeks or months to be exploited in infrastructure — nowadays that's minutes, if not seconds. The speed at which your infrastructure and your operational teams can react to changes in vulnerabilities is something that AI absolutely can be used to assist your business." — Shane Carlson

Not everywhere, not yet

Even those who haven't seen it firsthand can see where this is going.

"I see some early winners in this space that are actually taking advantage of things like generative AI and other automated capabilities. Me personally, I've not seen it yet." — Doug Tedder, IT service management and governance consultant

The early winners are real, and what they are doing is working. The use cases in this piece are already reducing noise, closing incidents, routing tickets, and freeing up the people who used to do all of that manually. The organizations doing this well aren't waiting to see what AI becomes. They're already living with what it is. And what it is, right now, is genuinely useful.