The Question Nobody Answers
I asked her a simple question: if AI did the work instead of her, how would that even get billed? She didn't have an answer straight away. She works for a small R&D company funded largely through public grants, and after a moment she landed on one I liked: it would be like broadband, you'd just pay an averaged slice of the subscription. The thinking itself wouldn't be worth much on its own at all.
Then I asked the follow-up. What if a skilled person used AI to do their existing job in a fraction of the time? Same answer, roughly: it wouldn't change what they're paid. It's just a tool, like learning to use a spreadsheet properly. That's the part that's never sat right with me. If a tool makes someone faster, the time it saves has to go somewhere. The only real question is who gets to keep it.
Changing What Winning Looks Like
In an ideal version of this, we don't just fix the pricing model, we change what winning looks like. Businesses that want to attract new talent will usually pay whatever the going rate demands, that's just how hiring works.
The outcome I actually want isn't complicated. The employee who becomes more efficient gets rewarded for it. The business benefits too. The client or consumer gets better outcomes faster, maybe even lower costs along the way. Nobody has to lose for this to work.
But an employee who becomes more valuable on their own initiative, not because the company trained them, quietly becomes a hidden profit nobody's tracking. Retention often sits with a completely different budget and team than recruitment, so there's rarely anyone whose job it is to notice, let alone reward, that kind of growth.
Treating it purely as a labour cost to be minimised is a false economy. Recruitment is expensive and finding the right person is one of the hardest parts of running a business, so keeping someone who's already good and getting better is usually the cheaper option, even if it rarely gets budgeted that way.
The Youth Footballer Problem
I like the youth footballer version of this comparison. A club pays a young player very little. If that player develops into a real asset, they often have to leave to get paid what they're actually worth. The club invested heavily in their development.
Where the analogy breaks down is what happens next: the club usually gets reimbursed with a transfer fee, so the investment pays for itself either way. Most employers get nothing back when a self-improved employee leaves.
That distinction matters. A company reasonably feels entitled to some return on skills it paid to build. But it has a much weaker claim over skills someone builds on their own time, which is exactly the position most people learning AI properly are in right now.
Whose Skills Are They Anyway
I'll put myself in that position. I'm someone on the sharper edge of learning how to actually use AI well, mostly on my own time, not because anyone funded it. Whatever value that creates isn't going to come back to me automatically, and I don't think that's fair, not just for me, but for all of us.
Collectively, we've all had a hand in training these models, typing corrections into a chat box, solving a reCAPTCHA, sharing code in the open that's now training the next generation of coding tools to build themselves faster still, at an exponential rate. Developers may well be training their own eventual replacements.
If we all had a hand in building this, we should all have a stake in what it gives back, and more time, not just more profit for whoever owns the model, is one of the fairest ways that could look.
Pay Isn't the Only Lever
We default to full time employment because it's the template we inherited, not because it's the only one that works. A team could reward finishing ahead of schedule with an extra day off instead of just quietly moving on to the next deadline. More roles could genuinely be part time.
That's not a soft benefit, it's parents picking their own children up from school instead of paying someone else to, and families with enough time and money left over to actually go and spend it together.
An economy with more of that isn't a smaller economy, it's a healthier one, and it matters more now than it has in a long time, because how we handle this says a lot about where we take things from here.
A Small Example, Because It's Honest
A conversation with a friend led me somewhere that writes its rules down properly. Publicly funded R&D is about as honest an example as you'll find.
UKRI, which Innovate UK sits under, publishes its own formula for what a person's time is worth on a grant: gross salary divided by working days, minus bank holidays and annual leave. That's the day rate. Non-PAYE staff are capped at £22 an hour.
It's audited, timesheets are required, a project manager has to sign off the hours.
But look at what the formula actually contains: hours and a rate. Nothing in it asks whether the work was hard, whether it took real skill, or whether AI just did in twenty minutes what used to take two days. The money moves through a hole shaped like "time logged," because time logged is the only thing anyone agreed was easy to check.
There's a principle in manufacturing called Kaizen, constant small improvement, and it only works if getting better is rewarded. Price purely by the hour and you've built a system that quietly punishes exactly that.
Innovate UK's project grants are released against agreed milestones, not paid out hour by hour. But the cost claim inside each milestone still runs on that same day-rate formula. Compress the work with AI, and the claim shrinks along with it, even though the outcome arrived exactly as promised.
The Same Friction Is Already Showing Up
It's not just Innovate UK grants either. R&D tax credits work on a similar assumption, and the cracks are already showing there too.
HMRC's official position on AI and R&D tax credits is neutral:
Using AI is neither sufficient to establish qualifying R&D nor a reason to deny relief.
Basically, "we're not sure yet." Fair enough, that's a genuinely hard line to draw.
James O'Leary at Kreston Reeves has gone further in public commentary:
What may have been a project qualifying for R&D tax relief one or two years ago could now be considered routine, and certainly the use of AI or related technologies will not necessarily mean that a project will qualify for the relief.
In other words: the project doesn't have to change at all for it to stop qualifying. AI raises the bar for what counts as genuinely hard, and the ground shifts under work that hasn't moved.
When what you're compensating people for was really "the hours it took to struggle through a hard problem," and AI quietly removes most of the struggle, every scheme built on that assumption starts to creak: a government grant, a tax credit, an agency invoice, take your pick.
Who Actually Keeps It
The real question sits a level above all of it, and it's not really about Innovate UK. It's one clean, well-documented example of a much bigger pattern, I've written before about how hourly pricing punishes the people who could get you there faster.
Say we fix the pricing model. Say the skilled person using AI gets properly rewarded for speed instead of penalised for it. Who actually captures that saved time?
Right now the honest answer is usually the business, the funder, or the client, as pure margin. That's a story about one side pocketing the difference, not a shared one.
I've argued elsewhere that the entire trajectory of this technology depends on who ends up holding the gains, whether AI's efficiency turns into cheaper outcomes and more leisure time for everyone, or just faster extraction for whoever already owns the tools.
Fixing how we price an hour of AI-augmented work is a small, practical piece of that much larger question.
Who keeps the time is.