Editorial
Emerging evidence suggests that AI is compromising our informal workplace networks. Left unhindered, this could lead to loss of peer trust, collaboration and social capital.
What can we do to arrest this harmful trend?
“Do you have a best friend at work?” is a question long used by Gallup in its annual employee engagement survey. Gallup claims it effectively differentiates high-performing teams from low-performing teams. Specifically, it serves as a proxy for trust: best-friend relationships indicate deeper levels of trust that boost engagement, productivity, safety and innovation.
The Rise of AI as a Best Friend at Work
The question came to mind when Pronita Mehrotra shared how her students were increasingly relying on AI for advice and support, rather than their class colleagues in a post on LinkedIn. She included the following graphic:
As a social/organizational network analysis (SNA) practitioner for more than 30 years, this graphic really caught my attention.
Like Gallup, we at Swoop Analytics had a favorite survey question: “Who do you rely on for helping you get your work done?” We typically asked respondents to pick their top five connections. From this single question we could create insightful network maps identifying who the “real” influencers are inside an organization.
Predictably, many of these influencers occupied managerial roles. However, invariably we would identify “hidden influencers” who managed to influence their peers from the bottom-up. These hidden influencers are often critical to any organizational change initiative.
So what were the attributes of a hidden influencer? These quiet achievers were ego-less; often feeling as if they were just doing their job. They were by nature very approachable and always aimed to please.
Beyond their nature, they also invariably knew ways of getting things done, including navigating around the constraining systems and processes in place. Sometimes it was a hard to use legacy IT system. Other times it was maneuvering around regulations or policies to produce a practical outcome. Some were long-term employees who just knew stuff — and if they didn’t know, they knew who would.
After seeing Mehrotra’s Before and After chart, I immediately knew that if ChatGPT was an option on a list of who I rely on to accomplish work, it would definitely get my vote. And I expect it is the same for many others. The resulting network map would indeed look like Mehrotra's “After” network.
The Rise of AI as Hidden Influencer?
AI mirrors many qualities of our traditional hidden influencers: ego-free, always available, adept at navigating complexity and remarkably nonjudgmental. What is not to like? Stories of AI becoming a ‘best friend’ at work are becoming increasingly common.
The Gallup engagement survey aims to provide a window into how engaged your employees are with their colleagues and hence, the organization as a whole. Organizations with cohesive friendship networks working for the common good, i.e. social capital, have increased retention, because people don't want to break those strong social ties and the performances they generate.
If AI becomes the hub of your workplace network, Mehrotra’s concern becomes real: peer-to-peer bonds dissolve, and the social glue holding high-performing teams together weakens.
From Trust Networks to Star Networks
Of course, AI is not just everyone’s best friend at your organization; it's also everyone’s best friend at every other organization. If our social networks do collapse into “star networks” with AI at the center and no social capital to bind employees together, what is left?
Are we moving to a world where AI-powered individuals will simply be competing with each other for their own personal gain? Are our hard-earned social capital assets now at risk?
What the Data Tells Us
Mehrotra's post mentioned the study by Hou et al, “All Roads Lead to ChatGPT”: How Generative AI is Eroding Social Interactions and Student Learning Communities, which conducted 17 semi-structured interviews with undergraduate computing students across seven R1 universities in North America.
We've observed a parallel downward trend in people-to-people engagement on Microsoft’s Viva Engage platform.
Across 73 organizations and almost three million employees, we observed:
- A consistent decline in “Engager personas,” those people who act as connection hubs.
- Fewer two-way relationships, suggesting a drop in reciprocal collaboration.
- An increase in @mentions, though likely shifting from dialog (“join in”) to broadcast (“FYI”).
These trends parallel the AI-help-seeking patterns found by Hou et al., suggesting a broader, systemic shift away from peer interaction. We have seen a similar, but not as distinct trend in Microsoft Teams conversations.
| 5 Signals Social Capital Is Declining in Your Organization |
|---|
| Fewer written replies to discussion forum posts |
| Shorter discussion threads |
| Lower active participation rates in ESN |
| Increasing online meeting no-shows |
| Decreasing chat and channel interactions in Teams environments |
The data is telling us that the people-to-people depth of connections is reducing over time. We have no evidence that AI is responsible. But we can speculate that it has likely been a contributing factor over the past year or two.
Conference
Sep
16
HR Tech Conference Las Vegas 2025
Register
Conference
Oct
27
Gartner HR Symposium/Xpo Orlando 2025
Register
How to Rebuild People-to-People Connections
Exploring generative AI is still very much being conducted in private by many. People will undertake private research before being confident enough to engage with a human colleague on how to better use AI tools. We can, however, take some purposeful actions to ensure we don’t lose the ability to connect on a human level:
- Treat AI as a first step in a collaborative activity. For example, if your team is charged with developing a new service or product, accept that step one is individual AI use to augment and develop their personal perspectives. But there is always a next step; where colleagues come together to share and develop the product or service going forward.
- Use your Enterprise Social Network (ESN) to host one or more AI communities. The purpose is to develop “best practice” in the use of AI. For example, one community might focus on vibe coding, where non-developer application builders can meet with their developer colleagues and discuss issues arising from its use. Another could be for intranet content editors to share how best to use AI to help write engaging news articles.
- Consider using generative AI copilots to help you find and develop new connections. In the Microsoft environment, the Microsoft graph is a rich source of network intelligence that can be exploited by AI. I have published previously about how to use AI to conduct organizational network analyses.
- Double down on the use of your ESN to focus on people-to-people connection events. “Ask me anything” forums; innovation hackathons; or information campaigns that solicit human connections, such as human-interest news posts. Use your ESN analytics to track and monitor people-to-people connections.
- Ensure that members participate in digital team meetings, rather than waiting for the AI summary. Use your Teams analytics to monitor in-person interactions.
Editor's Note: Read more about the unintended side effects of AI use in the workplace:
- The Quiet Politics of Admitting AI Use — AI use in the workplace is on the rise, yet few admit to using it. What can organizations do?
- The Double-Edged Sword of Using AI in Meetings — While AI can improve efficiency and inclusivity in meetings, it must be wielded responsibly. Some considerations to keep in mind.
- Are We Focusing on the Wrong AI Use Cases? — When we stop lumping all AI into one bucket, we can identify the different use cases worth focusing on (hint: they're not the ones you hear about the most).
Learn how you can join our contributor community.
