Installation Isn't Transformation
A Gartner finding has been circulating widely this month: 88% of HR leaders report no significant business value from their AI investments. Not that the tools failed. That the value never showed up. Those are two different problems, and it’s worth being precise about which one your organization actually has. It’s not about the software.
The finding has understandably struck a nerve. It’s being quoted, reposted, and rebuilt into commentary across the staffing and HR technology world — and for good reason. Anyone who has bought software recently has a version of this story sitting somewhere in their own operation. The pattern behind it is not — staffing firms have been living this exact failure mode since long before anyone said “agentic.”
The Adoption Problem Predates the Software
Here’s the part getting lost in the rush to talk about AI specifically: this isn’t an AI problem. Swap “AI” for “applicant tracking system,” “vendor management system,” “ERP conversion,” or any staffing back-office platform your firm has adopted over the last twenty years, and the argument doesn’t need to change a word. The failure pattern is identical. AI didn’t invent this. It just made the pattern visible again, faster — because the Gartner number happened to land this year.
The mechanism is almost always the same. A firm buys good software. Someone sends a launch email. There’s a training call half the team skips. Then everyone quietly goes back to the exact workflow they used before, because that workflow still technically works and nobody actually made the new tool the only way to get the task done. The tool becomes a thing sitting next to the job, instead of becoming part of the job. Multiply that across a staffing operation already running on thin coordinator and recruiter bandwidth, and adoption doesn’t so much fail as it never really starts.
It’s the same pattern the principals and consulting partners at Heagney Logan Group observed up close across decades of staffing technology change — ATS conversions, VMS and MSP program rollouts, ERP replacements, and payroll and billing system migrations that predate most of today’s platforms entirely. The technology changes. The failure mode doesn’t. A tool that gets installed but never gets adopted isn’t a technology failure. It’s an organizational one — and it was preventable before a single license got purchased.
The Gym Membership Problem
This isn’t a staffing-specific weakness, either. It’s a basic human pattern with follow-through. The New Year’s resolution that fades by the second week of January. The crash diet abandoned before it changes anything. The gym membership that outlives the actual gym-going by eleven months. Good intentions, real money committed, and no structure underneath any of it to turn the intention into a habit.
The gym membership is worth staying with for a moment, because it maps almost exactly onto what an unused AI seat looks like on a monthly software invoice. The intention was real. The investment was real. What never showed up was the structure that turns intention into a habit: a specific plan, someone checking whether it’s actually happening, and a fast correction when it isn’t.
A gym doesn’t actually need you to show up. The membership renews whether or not you ever walk through the door.
A trainer is a different relationship entirely — a trainer’s job depends on you getting stronger.
That distinction is the whole point. A good trainer checks whether you showed up, adjusts the plan when something isn’t working, and treats a missed week as information instead of a footnote. Most staffing firms buy the gym membership. Almost none of them hire the trainer. And unless someone is playing that role deliberately — with real authority, not just enthusiasm — the pattern above repeats itself with the next platform, and the one after that, because a team that has already quietly decided “that tool didn’t work” carries the verdict into the next purchase before it even arrives.
Why This Predates AI — And Why That’s the More Useful Framing
None of this is really about whether agentic AI specifically works. It clearly does — the technology is not the bottleneck here, and treating it as one wastes energy in the wrong direction. The more useful question isn’t “should we adopt AI.” It’s the same question that should have been asked before the last ATS migration, the last VMS integration, or the last payroll system conversion: is the workflow this tool is about to run actually documented, understood, and stable enough to hand off — to a person, a platform, or an autonomous agent?
Most firms skip straight to buying the tool and hoping. The step that gets skipped is the one where someone documents what the process actually is today, including every exception a coordinator or recruiter has learned to handle by instinct, before deciding what to automate and how.
What Documenting the Real Process Actually Looks Like
A useful diagnostic here doesn’t start with technology at all. It starts with documenting operational reality — independently, from the people actually doing the work, before anyone tries to summarize or fix anything. That reality gets examined across six consistent dimensions: People, Process, Technology, Measurement, Management, and Methods. Only once that picture is accurate and validated does it make sense to classify what’s actually broken. Only after that does a solution — AI-driven or otherwise — get evaluated.
Skipping that sequence is exactly how a firm ends up automating its own dysfunction faster, instead of fixing it.
The Real Question Isn’t Which Tool. It’s Which Kind of Problem You Have.
Turning on a new tool — AI or otherwise — inside an operation that’s already well-documented, centralized, and running on clean data is a straightforward installation. It works close to out of the box, because the organization underneath it was already ready.
Turning on the same tool to paper over an operation that runs on undocumented judgment calls, disconnected systems, and inconsistent process is a much bigger undertaking wearing a much smaller costume. It fails the way every under-scoped transformation fails: quietly, expensively, and with the technology absorbing the blame for a problem it didn’t create.
Knowing which one you’re actually looking at, before the purchase order goes out, is the entire game.
Start With the Process, Not the Platform
Before the next platform conversation — AI agent, ATS, VMS, or anything else — it’s worth answering one question honestly: is this specific workflow, in this specific business, documented and stable enough to automate? An independent, vendor-neutral read on that question, before any vendor is in the room, tends to be the difference between technology that sticks and technology that quietly joins the list of tools nobody uses anymore.
