Most AI pilots in local government don’t fail because the technology doesn’t work. It fails because no one changes the way the work is done around it.
This is the most useful thing we have learned during our three years of work. We launched Outcomes Matter Consulting in 2023 and have since been alongside caseworkers, social workers, SEND staff and service leaders on the ground as they seek to incorporate new tools into their already packed working day. The pattern is consistent. A license is purchased. Training is by reservation only. And after six weeks, half the team quietly went back to their old ways. Because the old way was the way it fit everything else they had to do.
This is not a workforce issue. That’s a design issue. Throwing a tool at a process that hasn’t changed is just one more thing to learn on a slow day. Humans have no resistance to AI. They resist being handed something that makes their day more difficult before it makes their day easier, and not being able to support the gap in between.
Therefore, important work is rarely technical work. What surrounds it is behavior and culture. What would a good day look like if we had this tool? Who would review the output and how would we build trust in that output? Where would the time saved actually go and who would decide? What happens the first time something goes wrong? None of these questions are answered by software. Everyone will decide by Christmas whether the software is still in use.
To be honest, adoption has been slow and should be the case in public services. These are the teams that make legal decisions regarding children and vulnerable adults. Attention doesn’t get in the way of design. My characteristic is that I work responsibly. The mistake is to treat careful implementation as a reason to do nothing rather than a reason to do it right. Governance built alongside the work, boundaries of human participants that the team can see and trust, and impact measured honestly rather than asserted.
Here’s what changes when you adjust the conditions, not just the kit. Social workers consistently say that about 80% of their time is spent on administration and only 20% on direct work that actually changes outcomes. This is the worst managerial ratio of any major profession in the UK, and it’s not the fault of those inside. This is what a system configured this way produces. What we can gain from AI is not new. That ratio has shifted, even slightly, so that more time is spent on the costliest and most human part of the job.
If you like this content…
But the prize will only be won if the time saved is protected, named and shown somewhere. We’ve seen teams spend hours on end, then quietly absorb that time into their backlog. Technology created the space. Space was wasted because there was no planning for the space. This is not a technology issue, but a behavior change failure that cannot be fixed by upgrading.
Therefore, the advice we offer after three years will be brief. Start with the actual pressure your team is already feeling, rather than a tool someone wants to try. Bring in your frontline troops early. Because it will tell you what is broken. Build governance in parallel rather than later. Also, since logging in is the easy part, treat the rollout as the beginning of your work rather than the end.
None of this is the exciting part. The demo doesn’t work. But it’s the difference between a pilot that quietly fades away and a change that’s still there a year later, doing real work for conversations that actually matter to families, and releasing in real time.
If you get the technology right, you have the tools. Change can occur if you create the right conditions. When I went back and looked, only one of them was still standing.
Because results matter.
