When we hire a graduate, we expect to invest in them. We train them. We give them time. We let them get things wrong, and we help them get better. Nobody hands a new starter a complex brief on their first morning, watches them fumble it and walks them straight back out of the building.
Yet that's exactly how a lot of organisations treat their AI tools.
I was talking to Yuliya Fontanetti, Chief Digital Transformation Officer at MMR Research, on the latest episode of The Next Thing Now, and she put this better than I've heard anyone put it. We're very happy to invest in people. With software, we try it once, it gets something wrong, and we decide we haven't got time for it.
I think that goes a long way to explaining the AI adoption problem.
Some of this is self-inflicted by the industry. The big vendors spent two years promising magic, somewhere between the end of the world and a shiny utopia with unicorns and rainbows. Quite a lot of that was marketing. Plenty of organisations paid for licences, got next to no ROI and came away suspicious.
That suspicion is understandable. The trouble is what happens next. One poor result becomes the verdict. The tool gets shelved, the pilot gets labelled a failure, and the organisation moves on convinced there's nothing in it.
Trust makes or breaks software, in the same way it makes or breaks a relationship. The difference is how easy it is to walk away from a tool.
What struck me about MMR's approach is how deliberately they've kept the door open. When a tool doesn't deliver, they park it and come back to it a month later. They keep tabs on it. They watch it change.
That sounds like a small thing. It has a big knock-on effect. The tools they looked at a few months ago are completely different today, so a tool that failed in the spring might be the right answer in the autumn. People inside the organisation learn that "not yet" is a legitimate answer.
The mood shifts from "this doesn't work, I'm never touching it again" to "I think it's got some legs". That's the R&D mindset. Try something, find what's good in it, come back when it's moved on.
Treating tools like new hires only works if you invest in the humans around them as well.
MMR ran a pilot group before rolling tools out across the company. A few months later, they went back and retrained that same group, because the tech had moved and they'd built agents the pilot group had never seen. That's a humbling thing to admit and a vital thing to do. The people who went first are the ones most likely to be working from an out-of-date picture.
They also set up an AI Task Force, with representation from every office and every level. It shapes onboarding and training, checks whether the agents they've built are still fit for purpose, and runs a proper feedback loop.
I loved what Yuliya said about positive feedback. We're all quite good at telling people what's broken. When something works, people just expect it and say nothing. She wants to hear about it: how much time was saved, whether the quality went up, what the use case was. That's how good practice spreads across a business.
You build trust in the good times, so you've got something to lean on when things get hard. If you're trying to build trust in the middle of a crisis, you're going to really struggle.
For a lot of people, AI feels like a crisis. It's being driven from the top, usually in the name of efficiency, with a background hum of "this is the end of work as we know it". Asking people to commit to a change that might not be good for them is a big ask.
MMR's leadership made space. Time to experiment, a trusted team to lead it, and permission for things to take a while. The top-down drive gave the bottom-up curiosity somewhere to grow.
A lot of leaders will find this one hard. We wouldn't ask a grad to justify their salary in week two, and the same patience applies here. The return arrives, often in places you least expect.
A colleague of Yuliya's in Colombia built a tool that could save a hundred days of work. That work had landed on the team as an unintended consequence of something else. Two years ago, clearing it would have meant hiring three or four people. Now the team gets that time back without eating into their evenings and weekends, and puts it into the work they're valued for.
That kind of return shows up in engagement, in brain space, in people doing better work. It's just as powerful as a line on the balance sheet, even if it takes longer to show up there.
Start with the problem. Tech for tech's sake solves nothing. Find the thing people find tedious, or time consuming, or properly hard. Then get your hands on the tools yourself, break a few things, and find the other people in the business who want to get going too.
When a tool lets you down the first time, give it the same patience you'd give a new starter. Revisit it. Retrain around it. Listen to what's working as closely as you listen to what's broken.
That investment is vital. It turns a failed pilot into a working capability. It builds trust. It's how the transformation sticks.