A developer I placed a couple of years ago sent me a link on the evening of 22nd October, with a shrugging emoji and no message.
It reminded me of a head of IT I know, who told me her team had got faster and her code had got worse. She had let her developer headcount slide by a couple of people, on the sensible theory that the AI tools would cover the gap. And output went up. What nobody had budgeted for was the person who has to read all of it.
Barclays' third-quarter results were behind that link.
The bank said it expects to cut about 5,000 jobs by the end of the year, roughly 5% of its workforce, with around 1,000 already gone and the rest due before December ends. About 60% of the cuts fall in support functions. And that group includes software developers, according to The Next Web's report on the results.
Developers are in that number.
What did Barclays actually say about developers?
According to digit.fyi's report on 23rd October, Barclays' finance director said the cuts include fewer developers, because AI is making the remaining staff more productive. The chief executive was more careful, tying the savings mainly to years of technology spending. Profit before tax still rose 12% to £2.3bn for the quarter, so these are not the cuts of a business in trouble.
I'm not going to pretend one bank sets the market.
But plenty of mid-sized firms watch the big names for permission. A board member who reads that headline on Friday will ask the IT director on Monday why the developer team looks so large, and that is a fair question. It just tends to arrive without the follow-up.
Why does cutting developers make the senior hire harder?
Because the work that remains is the work AI can't do alone. Reviewing generated code, spotting a bad architectural assumption and deciding what to build next all sit with experienced people, and in my experience those people are scarcer than the junior developers the tools replace. Take out a layer of developers and the senior engineers above them stop being a luxury and become the bottleneck.
And lots of firms are building a wide base and forgetting the people who hold it up.
I wrote last week about the government's push for entry-level tech roles, and the same hole in the middle showed up there.
Somebody has to lead all those new starters, and that somebody does not appear by magic. When I take a brief for a lead developer or an engineering manager at the moment, the job description has usually shifted too. Less build the platform. More own the quality of whatever ships, including what the machines wrote.
And nobody seems to have told the salary band.
But the role is tougher and it still carries last year's number, and candidates who can do it notice that quickly. They have options, after all, and the good ones tend to be talking to someone else by the time you call.
Who ends up checking what the AI wrote?
In my experience, the same small group of senior people who were already stretched. Robert Half research reported by ITPro on 30th September found that 38% of UK tech professionals now spend more time overseeing and checking AI-generated output, while 53% said AI had cut the time they spend on routine tasks. I suspect the time saved at the bottom of the team is being spent again further up.
That is the head of IT from earlier.
Her developers were quicker, honestly. But her two most senior people were drowning in pull requests, and one of them had started leaving work at six sharp, which in my experience is never a good sign. Then she asked me what a replacement for him would cost, and I had the awkward job of telling her.
The same research found 45% of UK tech professionals are expected to build new AI-related skills.
Which sounds fine until you remember that learning a skill and judging other people's use of it are different jobs. One takes a course. The other takes years of getting things wrong in production.
Is anyone still hiring in tech?
Yes, but for different shapes of role than a year ago. In the same Robert Half survey, 47% of UK employers said they plan to grow their tech workforce before the end of the year, with cyber security the most sought-after skill at 54%, followed by agentic AI at 50%, generative AI at 48% and cloud at 44%. A quarter are expanding contract and project-based hiring.
The contract number matters more than it looks.
A lot of firms are quietly deciding that review and oversight capacity is better bought in brief spells than carried permanently. Sometimes that is right. Other times it means paying twice, because the person who finally understands your codebase leaves the day the contract ends and takes the understanding with them.
But I'd be wary of anyone who tells you there is one correct answer, me included. It depends on how long the AI-heavy phase of your roadmap really lasts, and nobody knows that yet. Not the vendors, whatever their slides say. Nor me, though I get asked weekly.
A small test before your next headcount meeting
Ask one dull question. When AI tools write more of your code, who in the business is accountable for whether that code is any good?
If you can give a name, check that person has the time and the seniority to do it. Or check their pay reflects the job as it now stands. And if all you get is a pause, you have found the vacancy, even if it isn't on the careers page yet.
Decide early whether the role is permanent or contract.
Candidates at this level tend to be in demand, and a firm that spends six weeks arguing about which it wants will usually find the shortlist has gone elsewhere. I've watched it happen more than once. The managers involved were never short of good intentions.
My head of IT is hiring a lead developer now, not to write more code but to read it.
If you have a senior IT role and you're not sure what shape it should take, have a chat with our IT recruitment team whenever you're ready.
