Copy article

The Babysitting Tax: Why AI will not be a cheaper replacement of engineers.

ended 09. December 2025

The story that an AI is going to steal the engineering teams jobs in the middle of the night has now met a reality check, and it was written by the insiders of the house, namely, engineers at Microsoft and LinkedIn. The fact that they have acknowledged that AI coding agents are fragile and prone to hallucinations and need to be babysat is a confirmation to those who have known, long enough, that code is inexpensive, yet production is costly.

What we are experiencing is a perilous kind of automation theatre in the development of software. C-suites are being sold the illusion that they can cut the number of people and boost production. The fact, which this report confirms, is that we are merely passing the price of writing code to the cost of debugging it. You have not saved time when an agent hallucinates a security defect or uses an obsolete SDK since it no longer has enterprise context that you must now exert high-interest technical debt against in the future that a human expert must pay off.

This is not a failure of the technology but it is failure of expectations. We must cease to regard AI as a lesser developer alternative, but we must regard it as a tool that must be used by a master craftsman to be effective. In the event that you are not able to check the architecture you should not bother to automate the building.

We'd like your views:

  • Is the industry using code faster as a replacement of providing value sooner?
  • Who bears the liability in cases of breach of security by the code generated by agents, the vendor or the implementer?
  • Will we end up with a generation of lost junior developers that can prompt but not debug?
  • Is the babysitting price of existing AI agents in fact greater than the wage savings of team reduction?
  • What is your rationale on the ROI of a tool that raises the rollback rate of your production environment?
  • Is there a reason why AI-generated code should be explicitly prohibited in enterprise governance core security modules?

3 responses from the Newspage community

Copy all

Star Quote
Copy

Rohit Parmar-Mistry
Founder at Pattrn Data
Corporate memory is frighteningly short. Ten years ago, the C-suite strategy was to outsource everything to large consultancies to 'optimise costs.' The result? A decade of fighting to bring knowledge back in-house because product quality plummeted and innovation stalled.

Now, they are doing exactly the same thing with AI. They are treating intelligence as a commodity to be bought cheap, rather than a capability to be built. Firing your junior engineers to replace them with AI agents isn't innovation; it's asset stripping.

Just like the outsourcing boom, this will look great on the balance sheet for two quarters. But when the consumer experience degrades and the 'babysitting tax' of fixing robot-generated errors piles up, companies will panic-hire humans back at double the cost. It’s the same old story: sacrificing the product to appease the spreadsheet.
Copy

Editor
Kate Underwood
Founder & Chief People Strategist at Kate Underwood HR and Training
AI is not here to steal your engineers’ jobs, it is here to wreck their working lives if you use it badly. Right now too many leadership teams are chasing headcount savings and ignoring the human fallout. Seniors become permanent babysitters for glitchy agents. Juniors never learn to think, only to prompt. That is how you destroy morale, skills and your future talent pipeline in one go. When AI generated code triggers a breach, the regulator will not blame “the bot”. They will look for the accountable human and the training, policies and governance you did or did not put in place. My rule for you lovely lot: if you cut people because of AI, you must invest double in development, clear ownership and psychological safety, or do not be surprised when your best tech brains leave.
Copy

Colette Mason
AI Ethics Consultant at Clever Clogs AI
Right now, AI coding agents are brilliant copilots when you've got human oversight but they're catastrophic in autopilot. But there's another factor to consider. Just like cloud storage, when the AI service goes down, your entire development pipeline stops dead and there's nothing you can do about it. Plus the rest of your operations and customer-facing workflows will be sunk too. A refund on your monthly subscription doesn't compensate for missed deployments, broken builds, teams sat idle waiting for someone else's infrastructure to come back online, and a tarnished reputation.

Fancy AI marketing promises the world for pennies per session, but it's more important to consider the cost of outsourcing your ability to ship code on a platform whose uptime and accuracy you don't control.