Stay informed with free updates

The writer is an FT contributing editor and the author of ‘Inside the Leaders’ Club: How top companies deal with pressing business issues’

It is January, and people have been making their technology forecasts. Will this be the year that AI proves its workplace usefulness? Are AI agents going to arrange our meetings, book our business trips, generate marketing reports and train our recruits? 

I am not a digital native: when I started in journalism, we composed our articles on typewriters and sent them by telex (If you don’t know what a telex machine was, ask ChatGPT — or your grandparents). But nearly five decades of writing about the progression of desktop computers, mobile phones, laptops, smartphones and beyond have taught me a few things about what makes technology succeed — and fail — at work. 

First, if office technology is going to catch on, it must be comfortable. In the early 1990s, many companies were excited about virtual reality. Travel businesses enthused about people experiencing destinations before booking, salespeople thought VR would help customers design kitchens and bathrooms, and businesses looked forward to training staff using headsets. 

The leading equipment manufacturer was VPL, a Californian company that made a VR headset called (and I’m not making this up) the EyePhone. The problem was that headsets were ungainly, with a protuberant front strapped around your head. The FT’s assessment in 1991 was that VR equalled “media hyperbole”, VPL filed for bankruptcy and other endeavours — such as an IBM-backed “Project Elysium” reimagining VR for professional settings — also flopped.

Subsequent attempts, such as Apple’s Vision Pro, have suffered from similar pitfalls. Workplace adoption has been largely limited to long-standing areas such as aviation training, and the total VR headset market has fallen by an annual 14 per cent, according to Counterpoint Research.

Another requirement for successful tech adoption? The idea should be explainable in less than a minute. When Mark Zuckerberg told us about the metaverse in 2021, changing his company’s name from Facebook to Meta, he had a glimmer of purpose. We were mid-pandemic, holding meetings on screens, and the tech promised to help us feel like we were sitting around the same meeting room table.

But in Zuckerberg’s telling, there was more to the metaverse. We would be cartoon-like avatars. There were flying fish. The avatars went surfing. The launch video felt like having Covid: the feverish intensity, mental confusion and, as the film was 90 minutes long, dread that it might never end. There was also an ungainly headset involved. The technology did not catch on in the boardroom, and after major losses, the company now looks set to cut back its metaverse drive. Our online meetings are still on screens with little squares.

Perhaps most importantly, if tech is to make us more productive, it mustn’t make users look incompetent. In law, for example, digital tools have long been able to replace drudge work, such as searching for precedents or drafting contracts. When I wrote about technology disrupting the sector in 2016, the lawyers and entrepreneurs I spoke to were confident that a costly, sometimes inefficient industry would be upended by advances such as automated search systems and virtual assistants, making legal services cheaper for all. But, 10 years later, expensive law firms are still very much in demand — as are lawyers to do the mechanical work technology promised to solve. 

Now, in the age of AI, tech champions say things really are changing, and that junior lawyers soon won’t be needed. So far, however, the scale of transformation seems to be as limited as the last wave of automation. Junior lawyers’ salaries have soared. As my colleague Sarah O’Connor has pointed out: “You don’t normally see the price of something going up when demand for it is going down.” 

One of the problems here is that lawyers still have to carefully check AI hasn’t made things up. A decade ago, tightened regulation following the financial crisis meant companies were frightened of getting things wrong, and reluctant to replace trustworthy human services with untested bots. Recently, lawyers in the UK and the US have landed in trouble for citing cases that didn’t exist after using AI. 

And industries far beyond law face the same reliability problem. Those AI-generated reports need to be minutely fact-checked. They are also in danger of producing more verbiage than any office needs — AI “slop”, which not only creates more work for staff but could have a negative impact on a company’s reputation.

Generative AI appeared in everyday life in 2022. In its publicly available form it is not yet four years old. It will only succeed in the workplace if it can grow out of a four-year-old’s predilection for mixing facts with fantasy.