AI is making work lonelier for the people who use it most

Every office runs on small inefficiencies. You need the latest sales figure, so you ask the person who keeps it. You cannot find a file, so you ask the person who saved it. You want someone to read your first paragraph before it goes out, so you lean over and ask. None of these exchanges would survive an efficiency audit. All of them are now easy to automate.

That is much of the appeal of artificial intelligence. A chatbot answers the question without sighing. An agent produces the status update without a meeting. A frozen laptop can be coaxed back to life without troubling the colleague at the next desk. Small talk is quietly disappearing from the working day. Taken one at a time, these moments look trivial. Taken together, they may have been doing more for people than anyone realised.

Our survey of 5,021 full-time employees in ten markets hints at the cost. Among the most fluent users of AI, 47% said it had led to less meaningful interaction with colleagues. Among intermediate users the figure was 39%; among beginners, 22%. The people who have embraced the technology most keenly are the likeliest to notice what it has displaced—and they do not seem to like it. Two-thirds of respondents said they wanted to feel more connected at work.

The survey cannot show that AI causes isolation; roles, workloads and workplace culture all play a part. Nor does the effect run only one way. Some 45% of respondents said AI had improved their relationships with colleagues. That is less contradictory than it sounds. A manager freed from admin can spend more time with a new recruit. An employee who uses a chatbot to marshal her thoughts may find a difficult conversation easier to start. Preparation can make for a better conversation. It can also become a substitute for having one.

The appeal of an "easier" colleague

Asking a colleague a question does two jobs. It solves the problem, and it builds an understanding of who knows what and how they think. A chatbot does the first job well. It does not do the second at all.

Part of its appeal is that it never judges. Nobody looks inexperienced in front of a chatbot. Microsoft's 2025 Work Trend Index, found that about 17% of those who turn to AI rather than a colleague do so for fear of being judged. That is a warning for managers as much as for technology teams. If asking for help invites embarrassment or impatience, people will look for private answers, and no amount of exhortation to "connect more" will change that.

The cost is subtle. In the MyIQ survey, 71% of respondents say they feel more self-sufficient at work, but 46% say they now know less about how their colleagues think and approach problems, and 38% say newer employees have fewer natural chances to build relationships. A worker can become better at working alone without becoming better known to the team. That matters when the time comes to decide whose experience is relevant, whom to involve in a decision or who deserves an opportunity.

We recommend you reading:Employee motivation: why the UK lags, and what actually drives it

The work that colleagues inherit

The retreat is not always voluntary. Sometimes it is colleagues who point people towards the chatbot to save themselves time. That is reasonable when a specialist is asked for the same file for the fifth time. But a routine question sometimes conceals uncertainty about a decision, and someone sent away too quickly may never mention what they actually needed.

AI can also make work look finished before it is. A polished document may oblige its recipient to check the facts, rebuild the argument and supply the missing context. Researchers at BetterUp and Stanford's Social Media Lab call this "workslop". In their survey of 1,004 American desk workers, 40% said they had received some in the previous month, at a cost of nearly two hours per incident. About half of recipients thought less of the sender's creativity, capability and reliability; 42% saw them as less trustworthy.

The sender may have saved time. The team may not have. This is where a technology problem becomes a management one. Reward output without regard to its usefulness, and people will pass on unfinished work; count only the time saved, and the checking, fixing and frustration borne by everyone else disappear from the ledger. Efficiency should be measured across the team, not at the desk where the document was generated.

What should employers keep “human”?

Whether people connect at work depends partly on whether the work gives them reasons, and permission, to do so. Clear expectations about AI, and deliberate room for real conversation, can protect relationships that efficiency would otherwise erode.

Start with conversations that carry responsibility. A manager may use AI to prepare for a pay discussion, but still has to explain the decision, listen to objections and follow up. An employee asking about promotion needs more than a description of the process; they need someone who can act. 

The same goes for feedback, conflict and concerns about wellbeing. Regular check-ins and clear routes to ask for help make that access real, and how managers respond matters as much as whether they are available. Admitting uncertainty should never invite embarrassment.

Learning needs the same care. Pairing a new recruit with an experienced colleague on a real decision gives both a reason to share judgment. Discussing why a project changed direction teaches more than circulating the final slides. Drawing another team into the conversation builds ties beyond the usual circle. But learning has to be rewarded in practice, not just praised in principle. Otherwise the organisation is asking its most helpful employees to subsidise collaboration with their own time.

Accountability belongs in the same conversation. Anyone who submits AI-assisted work should understand it, check it and make sure it is usable. Teams should agree when significant AI involvement must be disclosed, especially where automated messages or recordings could affect trust.

Above all, employers need to decide what to do with the time AI saves. If every efficiency gain is simply converted into more output, relationships will be squeezed into whatever gaps remain. Setting aside time for mentoring and joint problem-solving shows what an organisation really values. Summoning everyone back to the office does not solve the problem on its own: people can share a building and still never talk. What matters more is whether employees feel safe asking for help, able to challenge decisions and sure of someone they can rely on.

The importance of “third places” and social wellness

Even a well-run workplace should not bear the whole burden of belonging. People also need relationships in which their job title is irrelevant. Ray Oldenburg, an American sociologist, called the settings for them "third places": neither home nor work, but the café, the library, the park, the sports club or the the fitness studio. 

A weekly class or a walking group gives people a shared activity and a reason to learn each other's names.

Social wellbeing is about the quality of those relationships: being known, having people to turn to and being able to give something back. A full diary does not guarantee it. Community also keeps good habits alive. In Wellhub's survey, 64% of employees mentioned social support as extremely or very important to sustaining wellness habits over the long term, and 41% say their wellness routine helps them feel connected to others, not just fit. 

We recommend you reading: Your team already uses AI for its mental wellbeing. Is that safe?

AI is no substitute. Only 20% of those who use it for wellbeing support think it adequately replaces human connection.

A membership is only a start. A gym can be anonymous; a regular class produces familiar faces. A solo wellbeing app may help with routines but creates no social contact. The question is whether an activity offers chances to interact, and whether it suits the person. Access also depends on time: 46% of employees say their company does not give them the time and space in the working day to look after their wellbeing. A benefit nobody has time to use helps nobody. Employers can lower costs, make schedules more flexible and offer options for different abilities, interests and caring responsibilities. Participation should always be voluntary, whether employees want to join colleagues, meet new people or spend time with their own community.

Wellhub widens those choices with access to gyms, studios, group classes and other wellbeing services. At companies that offer it, 71% of employees say wellness is genuinely built into how their organisation operates, compared with 32% of those with another programme. That does not prove the membership caused the difference. The opportunity for employers is to pair access with time and choice, so that shared routines have room to form.

Those small office inefficiencies were never really inefficiencies. They were how colleagues got to know one another. AI can save the time they took. Only people can decide to spend some of it on each other.