AI for IT companies has moved past the question of whether to adopt it. Nearly everyone already has. The harder question is why so few are getting anything back. If you run an IT services firm, a managed service provider, or an internal IT organization, you’ve probably bought the tools, run the pilots, and given your team access to AI assistants. And you’ve probably noticed the results don’t match the promise.
You’re not imagining it. The data backs you up.
The adoption gap nobody talks about
Here’s the uncomfortable truth. Nearly 80% of organizations report regular use of generative AI in at least one function, yet only 5.5% of organizations see real financial returns from their AI investments. A widely cited 2025 MIT study found that 95% of enterprise AI pilots deliver zero measurable return.
For IT companies specifically, this gap stings more, because IT is the function where AI adoption runs furthest ahead. McKinsey found that agent use is most commonly reported in IT and knowledge management, where use cases like service-desk management are already in play. So your industry isn’t behind on tools. It’s often ahead. And being ahead on tools while behind on results is a specific, fixable problem. McKinsey & Company
The reason most deployments stall has almost nothing to do with the technology. It’s organizational. McKinsey’s research, drawn from a survey of 1,993 companies, identified the single biggest driver of whether AI delivers earnings impact, and it wasn’t model quality or budget. Among 25 organizational attributes tested, the redesign of workflows had the biggest effect on an organization’s ability to see EBIT impact from generative AI. The catch: only 21% of companies using generative AI say they’ve fundamentally redesigned even some of their workflows. McKinsey & CompanyMcKinsey & Company
The other roughly 80% are doing something that feels productive but isn’t. They’re layering AI on top of processes that were never designed for it.
What “bolting AI on” looks like in an IT shop
Picture a mid-sized managed service provider. They handle help desk, infrastructure monitoring, and patch management for a few hundred client endpoints. Last year they bought an AI assistant for their support team and a separate AIOps tool for monitoring. Both are good products. Both work as advertised.
A ticket comes in. The AI assistant drafts a decent first reply, so the technician saves a few minutes. The monitoring tool flags an anomaly, so someone investigates faster. These are real gains. They’re also small, scattered, and invisible on the P&L.
Why? Because the underlying workflow never changed. A human still routes the ticket. A human still copies context from the CRM into the assistant. A human still decides whether the monitoring alert matters, then opens a separate ticket, then re-enters half the same information. The AI speeds up individual steps inside a process built for people doing everything by hand. It’s like installing a jet engine on a horse-drawn cart. The engine works fine. The vehicle just isn’t designed for it. Libertify
The high performers don’t work this way. They treat AI as a catalyst to transform their organizations, redesigning workflows and accelerating innovation. In a support context, that means rebuilding the runbook so an agent handles first-level resolution end to end and escalates only genuine edge cases, rather than handing a human a faster way to do the same manual dance. McKinsey & Company
Redesign the operating system, not the tasks
This is where the conversation about AI for IT companies needs to shift. The goal isn’t a faster help desk. It’s an operating system where AI agents, data, and people each do what they’re best at, and the handoffs between them are designed rather than improvised.
That redesign rests on two foundations most firms skip.
The first is clean, connected data. AI agents are only as capable as the information they can reach. If your ticketing system, monitoring stack, asset inventory, and client records live in separate silos with inconsistent formats, an agent can’t act across them. It can only assist inside one. Roughly 73% of organizations report data quality as their biggest challenge in AI implementation, and in IT environments stitched together from years of acquired tools, that challenge is acute. Before an agent can resolve a ticket end to end, the data it needs has to be reachable and trustworthy.
The second is human oversight built into the design, not added as an afterthought. McKinsey reports high performers are far more likely to have defined human-in-the-loop validation processes, 65% versus 23% for everyone else. For an IT company, that means deciding deliberately where an agent acts on its own, where it recommends and waits, and where a person must sign off. Get this right and automation feels accountable. Get it wrong and one bad automated change to a client environment erases months of trust. CX Today
This is the work we focus on at StrataBlue. We build AI-enabled operating systems for businesses using our BRAVE framework, which exists precisely because the gap between owning AI tools and getting value from them is an operations problem, not a software problem. The firms pulling ahead aren’t buying better models. They’re running better systems.
Where IT companies have a real edge
There’s good news in all this, and it’s specific to your industry. IT companies understand systems thinking better than almost anyone. You already know that a tool is only as good as the architecture around it. You’ve spent careers explaining to clients why their problem isn’t the new application, it’s the integration. That instinct is exactly what AI value requires.
The economic case is hard to ignore. One IDC analysis projects that every new dollar spent on AI solutions and services is expected to generate an additional $4.90 in the global economy, and IT and telecom are positioned to capture an outsized share of that as both builders and users. But that return doesn’t come from access to the technology. It comes from how deliberately you wire it into how work actually flows. Microsoft
The window for that advantage is narrowing in a particular way. As one synthesis of MIT, McKinsey, and Harvard research put it, first movers in technology adoption are being caught quickly, but first movers in organizational redesign are not. Anyone can buy the same agent you bought next quarter. Almost no one can quickly copy an operating model you’ve spent a year redesigning around your own clients, data, and delivery. Entrepreneurs’ Organization
Start with the workflow that hurts most
If you want to move from scattered efficiency gains to something that shows up in the numbers, don’t start by shopping for another tool. Start by picking the one workflow that costs you the most in time, errors, or client friction. Map how work moves through it today, including every manual handoff and every place data gets re-entered. Then redesign it as if AI agents were full participants from the start, with clean data feeding them and clear human checkpoints holding them accountable.
That’s a harder project than installing software. It’s also the one that separates the small group seeing returns from the large group still waiting. If you’d rather not navigate that redesign alone, the way we work with IT and services firms is built around exactly this problem.
AI for IT companies was never going to pay off through tools alone. It pays off when the system underneath them is built to let it.
