1 December 2026 · 18 min read

Work friendships: what a manager can and cannot do about them

A manager cannot make two people friends. A manager can decide whether anybody is left standing at the edge of the kitchen with a sandwich, and whether the person who gets the good project is the one they go for a drink with.

The table at the end of the kitchen

Nadia has nine people on her team, in a company of about ninety. At half past twelve, three of them, Ben, Carys and Joe, take the same end of the long table in the kitchen. They have done it for two years. The jokes are theirs, and some things in the team get settled there before they are settled anywhere else.

Tom started three weeks ago. He eats at his desk and says he is fine. On Thursday Carys tells a story that ends "well, you had to be there", and Tom laughs a beat late. On Friday Nadia and Carys go for a drink, as they have since before Nadia was promoted. On Monday Carys gets the project Tom had been quietly hoping for. Nobody has done anything wrong.

Nadia has two ideas and likes neither. One is a team night out, which will be Tom's third week of looking pleased to be there. The other is to say nothing, which is also a decision. The research has a better answer, and it starts with being straight about what she cannot do.

Nadia and her team are invented. What follows is not.

Do work friendships help a team?

Mostly, and the best evidence is about people staying, getting on and not being lonely. None of it shows that friendship does the work by itself, and some of it shows a price.

Start with a UK figure. In the CIPD's 2026 survey, 64% of UK employees who rated their relationship with their colleagues good said they were unlikely to quit, against 36% of those who rated it poor, and 73% were satisfied with their job, against 32%; for the relationship with a line manager the shares were 66% against 35% and 76% against 24% (CIPD, Good Work Index 2026: Survey report, July 2026, Figure 6).

Two limits. It measures a good relationship with colleagues, which falls well short of friendship, and the ratings are the employees' own, so it shows what goes together and not what causes what. The manager's figures sit beside the colleagues' and are, if anything, slightly wider.

Peers count in their own right. A meta-analysis of 161 independent samples and 77,954 employees found that coworker support and coworker antagonism were linked to role perceptions, work attitudes, withdrawal and effectiveness, and that the results held when leader influences were controlled for (Chiaburu and Harrison, Journal of Applied Psychology, 2008). The gloss is mine: a good manager does not cancel out the person at the next desk.

Then there is the cost of having nobody. In a time-lagged field study of 672 employees and their 114 supervisors in two organisations, greater workplace loneliness was related to lower job performance, and the two things that explained the link were that lonelier employees were less approachable and less committed to their organisations (Özçelik and Barsade, Academy of Management Journal, 2018). The authors start from the point that lonelier employees still want to connect. A lonely person is not a person who prefers it.

How common is it? In the CIPD's 2026 survey, 15% of UK employees said they always or often feel lonely at work, 28% sometimes and 57% rarely or never (CIPD, Good Work Index 2026: Survey report, July 2026, Figure 44). Most people are fine. In a team of nine that looked like the survey, about one would not be, and might well say, like Tom, that they are fine.

The famous figure, and what it says

There is one figure you have probably met already. Gallup reported in 2010 that just 30% of employees had a best friend at work, that those who did were seven times as likely to be engaged in their jobs, and that those without one had a 1 in 12 chance of being engaged (Rath and Harter, Gallup Business Journal, 19 August 2010).

Read it for what it is: Gallup's own measure of engagement, sixteen years old, an association in Gallup's own data, with no year or country named on the page. Gallup's point is the word "best": its page says early research found a best friend a better predictor of outcomes than a friend, or even a good friend. So it is not a promise that any friendship you encourage will make someone seven times anything. It is a reason to take close ties at work seriously, and a reason not to try to manufacture them.

What are the downsides of friends at work?

The research that finds the benefits finds the costs in the same data, and any manager who has had a friend on the team already knows it.

In a study of 168 insurance company employees, the number of multiplex workplace friendships (friendships that also involve work-focused interaction) in a person's network was positively associated with supervisor ratings of their job performance, and also had a negative indirect effect through emotional exhaustion that was offset in part by more positive affect; a second study of 182 restaurant and retail sales employees found that such friendships helped job performance through trust and hurt it through the difficulty of maintaining them (Methot, LePine, Podsakoff and Christian, Personnel Psychology, 2016). The authors call it a mixed blessing: colleagues who were also friends helped people do the job through trust, and wore them down through the effort of keeping the friendships up.

Why would a friendship at work cost something a friendship outside it does not? One answer is that a job is a different kind of place. Pillemer and Rothbard argued in a theoretical framework that four defining features of friendship (informality, voluntariness, communal norms and socioemotional goals) are in tension with four fundamental elements of organisational life (formal roles, involuntary constraints, exchange norms and instrumental goals), and that mutual self-disclosure and perceived similarity, which deepen friendships, also lead to downsides for individuals, groups and organisations (Pillemer and Rothbard, Academy of Management Review, 2018). That is a theory, not a measurement, and the verb is "argued".

In plain words, and the gloss is mine: friends give without keeping score, and an employer has to keep score, in pay, ratings and who gets the good project. A manager who is friends with someone on the team stands exactly where the two meet. Nadia has one foot in each, and so does Carys.

Warmth has a second cost, and a manager is less likely to notice it. Amy Edmondson, whose research is behind the idea of psychological safety, wrote that team psychological safety is not the same as group cohesiveness, which research has shown can reduce willingness to disagree (Edmondson, Administrative Science Quarterly, 1999). Again the gloss is mine: a close, friendly team can be a quiet one, because nobody wants to be the person who spoils it. A team can have plenty of friendship and very little candour.

Are cliques and gossip a friendship problem?

Partly. The same ties that make friends also carry gossip and leave people out, which is why a manager cannot treat them as separate problems.

On gossip, in a social network analysis, friendship ties between employees were positively related to both positive and negative gossip, while workflow ties were related only to positive gossip; the more embedded a pair were in the friendship network, the more likely they were to gossip negatively (Grosser, Lopez-Kidwell and Labianca, Group & Organization Management, 2010). Plainly: gossip travels along friendship lines, the unkind kind included.

You might conclude that gossip should be stopped. Here is why a ban is not a plan. In a longitudinal study of one organisation's complete network, observed three times, gossip between two individuals increased the likelihood that they would later become friends, while individuals with disproportionately high gossip activity had fewer friends in the network (Ellwardt, Steglich and Wittek, Social Networks, 2012). The authors define gossip as informal evaluative talking about absent colleagues. In that organisation it was one of the ways friendships began, and the heaviest gossips had fewer friends.

So the manager's lever is not the talking. Using case studies of cattle ranchers, a competitive rowing team and airline company employees, Kniffin and Wilson found that workplace gossip can serve positive functions when organisational rewards are fairly allocated at the level of small groups rather than at the level of individuals within groups (Kniffin and Wilson, Group & Organization Management, 2010). Three case studies, so "can" and not "does". The gloss worth keeping is that how the good things in a team are shared out has a bearing on what people say about each other.

When a friendship group becomes a clique

A clique is a friendship group seen from outside. The warmth is the same. What changes is whether it is also the only way into information and good work. Things a manager can see this week:

• One table, one set of jokes, and a new starter nobody has explained them to.
• Decisions made before the meeting, by people who had lunch.
• The same two or three names on the good work, and nobody who can say how they were chosen.
• Gossip that reaches you late and never reaches the person it is about.
• A person who has stopped offering ideas in the room where their colleagues finish each other's sentences.

Leaders are part of this. A meta-analysis of workplace ostracism found leadership characteristics to be its strongest related antecedents, ahead of personality and of contextual factors such as social support, and found ostracism strongly related to performance outcomes, well-being and organisational perceptions such as job satisfaction; the authors also found that the outcomes measured before ostracism were as strongly related to it as those measured after, so some outcomes may instead be antecedents (Howard, Cogswell and Smith, Journal of Applied Psychology, 2020). "Related", not "causes": the authors' own last finding is that the direction is uncertain, and some of what looks like the result of being left out may be a reason people were. The part to take seriously is the first.

What a manager can do about friends at work

A manager can change the conditions under which friendships form, and the fairness of what happens around them. It is a short list, and it is enough to matter.

Give people time and work to do together

In interviews with 38 people, representing 19 peer friendships, Sias and Cahill found that the move from coworker to friend was perceived as caused mainly by working together in close proximity, sharing common ground and socialising outside the organisation (Sias and Cahill, Western Journal of Communication, 1998). These are interviews from 1998 about what people believed caused it, but the ingredients are plain, and a manager has most say over the first: who works near whom, and on what. Tom and Carys will not become friends by being in the same company. They might if they are on the same piece of work for a fortnight.

Distance takes the first ingredient away. In the CIPD's 2026 survey, 80% of UK employees rated their relationship with colleagues in their team good or very good and 2% poor or very poor; excluding the self-employed, 10% of those working fully from home said they had no relationship with their colleagues, against 3% of those not working from home at all (CIPD, Good Work Index 2026: Survey report, July 2026, Figure 36). Most people rate their team well. A team at a distance has to have the proximity put in on purpose, on the calendar, because it will not happen in the corridor.

Make the table big enough for everybody

In field research in the firehouses of a large city, using a mix of qualitative and quantitative methods, Kniffin and colleagues found a significant positive association between eating together and work-group performance (Kniffin, Wansink, Devine and Sobal, Human Performance, 2015). One city's fire service, and an association, so no promises about lunch. But it is the opposite of a team-building day: something ordinary, repeated and part of the working week.

Gallup's authors are direct about the limits of a gesture. Gallup's Jane Miller and Amy Adkins wrote in 2016 that managers do not need to force relationships between team members but should create opportunities for people to get to know one another, and that encouraging friendships at work does not mean installing game rooms and holding weekly happy hours (Miller and Adkins, Gallup Business Journal, 14 December 2016). The test of an opportunity is whether Tom can take it without being the new person making an effort. An invitation that needs a drink or an evening is not open to everybody. An open one is easy to tell: nobody has to be brave to accept it.

Watch your own friendships

The team can see where a manager's time goes. At a large financial institution, employees who had more face-to-face interaction with their managers were promoted at a higher rate, and the authors say this mechanism could explain a third of the gender gap in promotions at that firm (Cullen and Perez-Truglia, American Economic Review, 2023). One firm, and the measure is time with a manager, not friendship. The authors describe the setting as employees and managers who take breaks together and talk about family and hobbies. A manager can look at their own week and count who gets that time. What Nadia can do about the drink with Carys is nothing at all. What she can do about the project is say, in a sentence, why she chose who she chose.

Treat being left out as a work matter

HSE's Management Standards include relationships, which HSE says "includes promoting positive working to avoid conflict and dealing with unacceptable behaviour" (HSE, What are the Management Standards?). Our guide to HSE's six areas has the early signs. Being left out is not always unacceptable behaviour, and it is always something a manager can see. Noticing that a colleague is no longer included in the lunch, the thread or the joke is part of a manager's job, and so is saying something about it once, privately, about the behaviour.

What a manager cannot do

The honest list is as useful as the first one.

• Make two people friends. Voluntariness is one of the defining features of friendship in Pillemer and Rothbard's framework, and Gallup's authors say managers do not need to force relationships between team members. A friendship that is required is a rota. Sias and Cahill found that the later steps, to close and almost best friend, went with life events, personal and work problems and the passage of time, and none of those is in a manager's gift.
• Ban them. Gallup's authors call a strict policy against friendships misguided. For gossip, one study of one organisation found that talk about absent colleagues made two people likelier to become friends. A ban would not end it. It would move it out of sight, which is where a manager is least able to hear it.
• Stand in for the people around the person. The Chiaburu and Harrison result held with leaders controlled for. A manager who gets on well with everybody has done their half and left the other half of the person's working day exactly where it was.
• Know who is lonely. The CIPD's 15% do not announce themselves, and in Özçelik and Barsade's study lonelier people were less approachable. Asking is the only way, and some people will say they are fine, as Tom did.
• Be everyone's friend and the person who decides. That is the tension at the centre of Pillemer and Rothbard's argument, and it does not go away when the friendship is warm and the manager is fair.

How MeaningFuel helps with work friendships

The daily insight: Inclusivity, Teamwork and Collaboration

A company chooses the values and skills it wants to work on, and each working day everyone gets one short insight on one of them, drawn from something a real person said. Work friendships are not a problem a company can choose for its daily culture: that is what a module is for. What a company can choose is what friendly, working teams run on: Inclusivity and Teamwork as values, and Collaboration as a skill. In an ordinary week that might put a line about who is in the room in front of Nadia's team on a Wednesday, and one about helping on the Thursday. It will not seat Tom at the table. It keeps the question of who is at the table in the week where the table is.

The module: Working Well Together

MeaningFuel has no module called work friendships, and a module cannot make friends. The nearest is Working Well Together, for a team whose people work side by side rather than together. A module is a short course of ten daily lessons across two working weeks. Booked for a team, it invites the people on it, and those who join get one lesson each working day.

Its own text says what it works on: "Respect is built through what people do for each other every day, not through how warmly they feel about each other." The first lesson names a state it calls parallel play, a child-development term for children playing next to each other rather than with each other, and says it is lonelier than solo work, because the others are right there. The team chooses three ground rules of its own in the first week and practises them daily. One lesson is the offer: one specific, bounded offer of help, what and until when, because "anything anytime" is a greeting card and "reviews until 3" is something somebody can take. Another is credit, precisely: when something goes well, name who did which part, including whoever did the joining-up. It borrows an idea from CREW, a programme from the US Veterans Health Administration in which each work group sets its own goals.

None of that is friendship, and the module does not claim to make any. In Sias and Cahill's interviews, working together in close proximity was one of the main things people credited for turning a colleague into a friend. These lessons are practice at working together.

What leaders read, and what they cannot

Insights shows how much of a team takes part and how the insights land, pooled, with nobody named. It is not a reading of friendship. It cannot show who sits alone, who is in which group, or who was left out of lunch. During a module the lead sees whether the team is doing it, and never anyone's answers or what they write. If Nadia wants to know how Tom is, she still has to ask Tom.

What it does not do

Neither the insight nor the module makes Tom a friend, and neither promises a result. The module says of itself that where collaboration fails along one break between two people, that pair needs a repair conversation, which it cannot replace. What it does is put offering, asking and crediting into a team's week on a schedule, so they do not depend on one good Monday.

Questions managers ask about friends at work

Should a manager be friends with their team?

A manager can be friendly with everyone and should be careful about being close to one or two. Friendship runs on giving without counting, and a manager has to count, in ratings, work and who gets the chance. If you have a friend on your team, say plainly how you decide things, and spend your visible time on the whole team.

How do you stop cliques forming at work?

You cannot stop friends forming, and you do not need to: stop a friendship group being the only way into information and good work. Put decisions where everyone can see them, rotate who works with whom, and offer the lunch, the thread and the project to the whole team. In one meta-analysis, leadership characteristics were the strongest related antecedents of exclusion, so start with your own.

Should managers stop gossip at work?

Not by banning it. In one organisation's network, gossip between two people made them likelier to become friends, so a ban would also cut a way people connect. A manager can refuse to be a channel for it, say what they heard in the place it is about, and share credit by group rather than by contest.

Do remote teams need work friendships?

They need the chance of them. In the CIPD's survey, 10% of those fully at home, excluding the self-employed, said they had no relationship with their colleagues, against 3% of those not at home at all. In the research above, friendships began with working together near one another. Remote teams need that put in the diary, on shared work, with an open invitation.

Back to the kitchen

Nadia cannot make Tom a friend. On Monday she can put him on a piece of work with Carys and Joe for a fortnight. She can ask Ben and the others to open the lunch, and sit at it herself. She can say why Carys got the project, in a sentence, to the person who did not. Then she can leave a seat free that Tom does not have to be brave to take.

• Work Friendships
• Managing Teams
• Gossip
• Inclusion
• Workplace Culture