Two-thirds of mid-market firms are paying for AI tools that nobody's using twelve months after purchase. That's not a technology problem. That's a budgeting decision dressed up as patience.
I know this because I keep having the same conversation. It usually starts with a CTO or operations director mentioning, almost as an aside, that they've got "a couple of AI tools that haven't really landed yet." I ask how long they've been paying the licence. North of a year, almost always. I ask what adoption looks like. Single digits - three people out of twenty, maybe five on a good month. And then I ask why they haven't pulled the plug.
We've already paid for the annual licence. We can't just walk away from it. We need to get something out of it.
I get it. That instinct makes complete emotional sense. But it's costing you real money, and I want to walk through why - and what to actually do about it.
The licence fee you've already paid is gone. Not flippantly - as a financial fact. Whether you use the tool every day for the next six months or never open it again, that money has left your account. So the only question that matters now is: does the expected future value from this tool exceed the future cost of keeping it?
Future cost isn't just the next renewal. It's the internal resource spent trying to make adoption happen. Every hour someone spends trying to coax colleagues into using a tool they've already rejected is an hour not spent on something that might actually move the needle. And there's a quieter cost too: every month a tool sits unused, the next AI investment becomes a harder sell internally. Your credibility for these decisions erodes, slowly and almost invisibly, until someone in a board meeting says "well, what happened to the last one?"
I'm not saying cutting your losses is always the right call. Sometimes it absolutely isn't. But the decision to keep paying should be made on the same financial logic you'd apply to any other operational cost that isn't delivering. If you wouldn't sign up for this tool today, knowing what you now know, then "we've already paid for it" isn't a reason to keep paying. It's a reason you feel bad about stopping.
Feeling bad is fine. It's human. But it's not a line item.
The biggest mistake I see firms make is jumping straight to a decision - keep it or kill it - without understanding why adoption is low. Because the "why" completely changes what the right response is.
Over the years, I've seen low AI adoption fall into four categories. They're not equally common, and they're not equally fixable.
Tool fit is the one that stings most, because it usually means the original purchase decision was flawed. The tool solves a problem your team doesn't actually have, or it solves a real problem in a way that creates more friction than it removes. I was talking to an operations director at a consulting firm last year - she'd bought an AI-powered document summarisation tool for her research team, and it was a perfectly good product. But the research team's problem wasn't that they couldn't summarise documents quickly enough. It was that they couldn't find the right documents in the first place. The tool was answering a question nobody was asking. She knew it within about six weeks. She kept paying for another ten months.
Implementation quality is probably the most common one I see, and it's the most fixable. The tool is genuinely the right fit, but it was deployed without adequate configuration, integration, or workflow design. Someone signs up for a trial, it goes well in a demo environment, and then it gets rolled out to the full team without anyone thinking about how it plugs into existing processes. The tool works. The deployment doesn't. I've seen this happen with tools that were genuinely excellent - wasted because the rollout was basically an email saying "here's a new login, have a play."
Change management failure is the one firms are least willing to admit to. The tool was deployed without adequate explanation, training, or support. People don't use it because they don't understand it, don't trust it, or - honestly - because nobody with enough seniority told them it mattered. You can have the right tool, properly implemented, and still get near-zero adoption if the message from leadership is ambiguous. I've watched this happen. It's painful, because it's so avoidable.
Use case mismatch is subtler. The tool is technically capable and properly deployed, but it's pointed at the wrong task. The technology isn't wrong. The application is. A document processing tool designed for high-volume, repetitive work being used in a boutique advisory firm that handles thirty complex, bespoke matters a year - that's a use case mismatch. The tool does what it says. It just doesn't do what you need.
Take ten minutes and be honest about which category you're in. Write it down. Because what you do next depends entirely on this.
Fix the adoption. This is the right move when the diagnosis is implementation quality or change management failure. The tool is sound - the rollout wasn't. Fixing it means a focused sprint, not a programme. Reconfigure the integration so it fits the team's actual workflow. Run proper training - not a webinar, actual hands-on sessions where people use the tool on their real work. And get genuine senior sponsorship. Not "the CEO mentioned it in a town hall." I mean someone with authority who will check in on adoption weekly and hold people accountable.
If you can't get that senior sponsorship, be honest with yourself about whether the fix will stick. Without it, you'll spend another three months and end up exactly where you are now.
Renegotiate the contract. This works when the tool has partial value. Maybe it's brilliant for five people on one team but useless for the other fifteen it was licensed for. Your negotiating position is straightforward: you commit to the narrower use case, and the pricing needs to reflect the actual scope. Most vendors would rather keep you at a lower tier than lose you entirely, especially if you can articulate exactly which use case is working. I've seen firms cut licence costs by 40-60% with this conversation. Vendors aren't stupid - they know what their usage dashboards say.
Cut your losses. This is the right call when the diagnosis is tool fit or use case mismatch. The tool isn't right for your firm. No amount of better training or workflow redesign will change that.
Before you terminate, check your contract. Most enterprise AI contracts have renewal terms, termination notice windows, and data portability provisions that can catch you out. I had a client - a financial services firm, about 200 people - who tried to cancel a tool that three people were using. Turned out they'd missed a 90-day notice window by about two weeks. Locked in for another year. The operations director was furious, mostly at herself, because she'd been meaning to cancel it for months and kept deferring the decision. The deferral cost her the window. Check the contract and act accordingly - don't let this become another reason to delay.
One more thing: if the AI vendor is also a broader technology partner you value, the termination conversation has dimensions beyond the specific tool. You can be direct about the tool not working without torching a relationship. Handle it carefully.
This is where it gets politically tricky. Telling your board that you're recommending the firm stop paying for an AI tool it invested in nine months ago can feel like admitting failure. In a lot of organisations, that makes the conversation harder than it needs to be.
The framing that works - and I've helped a few clients navigate this exact conversation - goes something like this:
Given what we knew at the time, the investment was reasonable. We evaluated the tool, it addressed a genuine need, and the business case was sound. What we've learned since deployment is that it doesn't fit our specific environment - our workflows, our team's operating patterns, our data structures. We couldn't have known that without trying it. And now that we do know it, continuing to pay for something that isn't working isn't due diligence. It's just continuing to pay for something that isn't working.
That's not a failure story. That's a learning story.
The documented diagnosis - which category the adoption failure falls into, what specifically didn't work - is your evidence base for the conversation. It transforms "we wasted money" into "we ran a pilot, we learned something specific, and we're making a rational decision based on that learning." There's a difference. Most boards can see it, if you present it that way.
It helps to frame the cut as budget redeployment rather than budget loss. Unused licence costs in mid-market firms add up fast - across four to six abandoned tools, you're often looking at a meaningful chunk of the IT budget sitting completely idle. Firms that actively decommission unused tools tend to redeploy that capital into things that actually work. You're not writing off an investment. You're reallocating it.
Most firms skip this bit entirely. An adoption failure has produced specific, genuinely useful information about your AI readiness - and if you don't capture it before you move on, you'll make a suspiciously similar mistake next time.
A few questions worth actually writing down the answers to:
What was the binding constraint? Data quality - the tool needed clean, structured data and yours wasn't? Systems integration - it couldn't connect to your existing stack in any meaningful way? Cultural willingness - the team simply wasn't ready to change how they work? Be specific. "It just didn't work" isn't a diagnosis.
What would need to be true for a similar tool to succeed? Not in theory. In your firm, with your people, with your current systems. If the answer is "we'd need to clean up our CRM data first" or "we'd need a dedicated change management resource for the first 90 days," write that down. That's your prerequisites list for the next investment.
What evaluation criteria would have caught this earlier? Could you have run a smaller pilot with five users before committing to a full-team licence? Could you have tested integration with your existing workflow before signing? Build whatever the answer is into your procurement process.
The licence fee is gone. The learning doesn't have to be.
If you're reading this and recognising your own situation - a tool quietly draining budget while the usage dashboard gathers dust - the worst thing you can do is nothing. Inaction isn't neutral here. It costs you money every month.
Run the diagnosis. Work out which category you're dealing with. Make the decision - fix, renegotiate, or cut - based on what the diagnosis tells you, not on what you've already spent.
If this isn't just one tool but part of a broader pattern - a few AI investments that aren't landing, or a nagging sense that your firm's foundations aren't right for the AI initiatives you're planning - we offer a short AI portfolio review that produces a clear recommendation for each tool in the portfolio: keep, renegotiate, or cut, and why. Sometimes having someone from outside the organisation say "this isn't working and here's why" is exactly what's needed to unstick the decision.
Either way, stop paying for patience you can't afford.