KPMG finds workplace friendships boost engagement and turnover risk

Workplace friendship is now a performance issue

KPMG and CivicScience's Friends at Work 3.0 draws a sharper line than most culture surveys. The report, fielded April 17 to 22, 2026 among employed U.S. adults with sample sizes ranging from 558 to 1,354 depending on the question, splits employees into four groups: close personal friends, workplace friends, collegial relationships, and non-socializers. That distinction matters because the highest-performing group is not automatically the safest for retention.

Employees with close personal friends at work are the most engaged, productive, and trusted, but they are also the most likely to be looking for a new role in the next 12 months. For consultants, auditors, and advisory professionals at KPMG, that is not a soft culture finding. It is a warning that the people carrying the most responsibility, and often the most client pressure, may also be the easiest to lose.

The numbers show why managers should care

KPMG's 2026 data gives managers a concrete way to think about connection rather than a vague one. Among employees with close personal friends at work, 49% say they are always engaged, 87% say they can focus on skill development, 47% trust their company, 39% trust leadership, and 55% manage others. Those are not just friendship metrics. They map directly to how a firm runs staffing, coaching, and promotion pipelines.

The same group leads AI usage at about 35%, which is especially relevant in a Big Four environment where technical tasks are increasingly automated and the remaining work is more judgment-heavy. If the employees most open to new tools are also the ones most embedded in client delivery and team leadership, then workplace connection is doing more than making people feel good. It is shaping who adapts first, who carries change through the practice, and who becomes visible in the next promotion cycle.

Not all connection produces the same outcome

KPMG's own framing is useful because it refuses to flatten every relationship into one bucket. Collegial relationships, the professional middle ground that sits below friendship but above pure transaction, scored the highest overall wellbeing at 59.4 in the 2026 report. That suggests there is real value in steady, low-drama relationships that make it easier to get through busy season, long client days, and transformation work without burning out.

Non-socializers, by contrast, report weaker outcomes, including lower happiness, lower excitement, and less support for mental health. That matters in hybrid teams, where isolation can be easy to overlook until it shows up as slower collaboration, more brittle feedback loops, and a manager who only hears from someone when a problem has already escalated. The lesson is not that every team member needs a best friend. It is that some level of connection is tied to endurance, and different forms of connection produce different business results.

The friendship premium is real, and so is the turnover risk

This is not KPMG's first pass at the issue. The firm's inaugural Friends at Work survey, released in November 2024, found that 81% of workers considered workplace friendships highly important and 79% had at least one work friend. A year later, the argument got more explicit: KPMG said 57% of people would take a role paying 10% below market if it came with close work friendships rather than a job paying 10% above market without them. In effect, the firm put a 20% salary premium on workplace friendship.

The 2025 report also said workplace loneliness had reached 45%, nearly double the prior year, while 87% of workers said close work friendships were very important, up 6 percentage points since November 2024. Put together, the three surveys show a clear pattern. Connection has become more valuable, but it is not evenly distributed, and it does not always push in the same direction as retention.

That is the uncomfortable part for managers. The people with the strongest bonds are often the ones who feel the most engaged, and also the ones with the most outside options. In a firm where compensation is reviewed in cycles and careers are measured against partner-track milestones, the strongest internal ties can strengthen performance right up to the point they make someone more marketable elsewhere.

What KPMG leaders should do with the finding

The practical response is not more generic bonding. It is better-designed connection. Close friendships can boost trust and energy, but collegial relationships may be the more durable foundation for hybrid professional-services teams because they appear to support wellbeing without concentrating turnover risk in the same way. That means leaders should build habits that create frequent, useful contact, not just forced socializing.

A useful management playbook would look like this:

• Build recurring team routines around client work, coaching, and debriefs, so connection is tied to delivery, not just after-hours events.

• Use mentoring and peer-buddy systems to widen access to support, especially for newer staff who have not yet found their people.

• Watch the employees who are both highly engaged and highly trusted. KPMG's data suggests they may be the same people most likely to leave.

• Treat AI adoption as a relationship issue, not only a technology rollout. The 35% AI usage among workers with close personal friends at work suggests trust speeds up change.

• Make collegiality easier in hybrid teams by reducing friction in handoffs, feedback, and staffing communication. A stronger working bond can do more for retention than one-off morale efforts.

For KPMG, the larger takeaway is clear. Workplace connection is not valuable because it sounds nice on a culture slide. It is valuable when it improves collaboration, retention, manager effectiveness, and the willingness to take on new tools. The firm now has three years of data showing that not all workplace bonds work the same way, and the smartest leaders will manage those differences instead of pretending every kind of connection produces the same result.