View / Companies have already lost control of workplace AI

Reed Albergotti
Reed Albergotti
Tech Editor, Semafor
May 6, 2026, 1:01pm EDT
Technology
Hooded man holds laptop computer as cyber code is projected on him in this illustration picture.
Kacper Pempel/Illustration/File Photo/Reuters
PostEmailWhatsapp
Title icon

Reed’s view

Corporate America is incinerating billions of dollars trying to build AI tools that white-collar workers actually want to use. Most of these efforts will fail.

The gap between what a well-funded IT department can offer and what an inspired intern can vibe code over a weekend is closing. Actually, the intern may already have won.

Shadow IT emerged 15 years ago when workers started using their iPhones and cloud applications for daily tasks and ditched the clunky internal networks their employers had spent a fortune building. Corporate IT departments tried to block these efforts, but failed. They eventually relented.

Today’s version is much more potent. The utility-to-risk ratio has shifted so far toward utility that companies have already lost control. Workers aren’t thinking about data governance, security, or the 100 other things that can go wrong when freelancing in software development. They just want to move fast — and now they have the tools to do it themselves.

AD

To take back control, companies have to give up on walled gardens and build gateways. The goal can’t be forcing employees into a one-size-fits-all coffin of enterprise software. It’s figuring out how a sea of vibe-coded apps can securely plug into corporate data, where new employees who arrive with their own software can interface with the company’s systems on day one.

Not only will that reduce the friction, allowing employees to work with tools that actually match how they think; it will also boost productivity and unlock a lot of latent human creativity in the process.

Title icon

Room for Disagreement

One rigorous real-world vibe coding in the workplace, a three-month trial of Microsoft’s M365 Copilot across 1,000 workers conducted in the UK, found no discernible gain in overall productivity, though users said they were broadly satisfied with the tool.

Title icon

Notable

  • Research by Morgan Stanley found that only 21% of S&P 500 companies could cite a measurable AI benefit at all, even as hyperscalers are on track to spend $675 billion on AI infrastructure in 2026.
AD
AD