More tools, more processes touched, same bottlenecks - just with an AI label on them now. This diagnostic helps you see where the friction actually lives.
Run the diagnosticFor the past several months, we've been talking with IT directors, ops leads, and engineering managers at mid-market software companies - teams of 5 to 40 people, companies ranging from 100 to 1,000 employees. The people who own this problem, not just the ones who approved the tool purchase.
What came up, across most of those conversations, wasn't that AI tools failed. It was that they got applied to the wrong layer of the problem. The fragmentation was already there. The coordination overhead was already there. AI landed on top of it, added new surface area, and left the root cause untouched.
After enough of those conversations - at SaaS companies, fintech teams, infrastructure orgs - the contributing factors became consistent enough to put into a short diagnostic. That's where this came from. Not a framework, not a benchmarking study. Just a pattern we kept hearing from people doing the actual work.
Most teams have tried at least one AI tool in the last two years. A copilot here, a search plugin there, maybe a chatbot wired into the help desk. Some of it helped. A lot of it didn't stick.
The pilots that worked tended to help one person, or one team, in one specific workflow. The company-wide rollouts mostly created new overhead - more tools to manage, more questions about which tool to use, more discrepancies between what the AI says and what actually happens.
The problem isn't AI itself. It's where and how it got applied. Most AI deployments got layered on top of broken processes rather than wired into the actual flow of work. The bottlenecks are still there. They're just harder to see now.
Problem 01
Even with AI search tools in place, employees still can't find the right answer reliably. Information lives across Confluence, Google Drive, Slack threads, and old email chains. AI tools didn't consolidate it - they added another search box to the pile.
IT and ops teams still field the same questions every week because people don't trust the tools to give them the right answer. And they're often right not to. The AI is only as good as the information it can reach, and that information is fragmented, out of date, or sitting in a system the tool doesn't connect to.
Problem 02
This one surprises people. AI tools lowered the threshold for asking questions. Employees who previously figured things out on their own now expect instant answers - and when the AI tool fails them, they escalate to IT anyway. The net result, for many teams, is more tickets, not fewer.
There's a second layer too: employees now submit tickets about the AI tools themselves. "Why did it give me wrong information?" "Where did my data go?" "How do I use this feature?" A whole category of IT requests that didn't exist two years ago.
The promise was self-serve. The reality is a new escalation path that didn't come with a budget for the extra headcount to support it.
Problem 03
Onboarding, access requests, approvals - these still require someone to chase someone. AI didn't get wired into the actual workflow, so the handoffs still happen over Slack DMs, emails, and tickets. Someone in IT or ops is still the connective tissue between systems that don't talk to each other.
The coordination cost is invisible until you start counting it. How many messages does it take to provision a new hire's software access? How many follow-ups happen before an approval comes through? That cost compounds every week, every quarter, and it's almost never in anyone's budget.
If any of that sounded familiar - the diagnostic puts a number on it. Takes 2 minutes.
Run the diagnostic5-question diagnostic
Please enter a number to continue.
Count everything - Slack, Confluence, Jira, Notion, Google Drive, email, wikis, ticketing systems, shared drives. Each one counts.
Please enter a number to continue.
Tickets, Slack DMs, email questions - anything that requires a team member to respond. Estimate is fine.
Please enter a number to continue.
Copilots, AI search, chatbots, automation tools - count each deployment, even pilots that didn't fully roll out.
Please select an option to continue.
Please select an option to continue.
Your estimate
-
hours lost per year to avoidable IT requests
The patterns driving this score - and where to focus first - are in the write-up below.
The number above is the surface. What's underneath - the specific friction pattern in your setup, what's driving it, and what teams in a similar position did first - that's what the write-up covers.
Three things came up consistently in your answers. I've written them up. If it's useful, leave your email and I'll send it over.
Please enter a valid email address.
Not a newsletter. A personal email with the write-up - that's it.
Got it. I'll send it over personally.
While you wait - our AI at Work 2026 research covers the broader data behind what this diagnostic is measuring. Worth a read if you have 10 minutes.