The research also points to a quiet but significant shift in workplace dynamics, with frequent AI users reporting lower levels of civility and social connection.

The findings raise important questions for leaders about how AI shapes the psychological and social conditions employees operate in, and what they can do to manage its influence.

“These tools present psychosocial hazards which are new and not on the WorkSafe list right now, and they also risk heightening other psychosocial hazards that we’re dealing with,” says McQuaid.

“The question for HR teams is, how can we make the invisible visible when it comes to the impact that AI is having on wellbeing and relationships? Because only then will leaders and technology advocates start to take that risk more seriously.”

Below, McQuaid unpacks what her research tells HR, managers and leaders about how AI is influencing wellbeing and workplace dynamics – and what organisations can do to respond.

How does AI impact employee wellbeing?

The sharp dip in wellbeing among frequent AI users can be attributed in part to our natural stress responses, says McQuaid. 

Many employees are learning to use new and fast-changing technologies on the job in real time, and face rising pressure from leaders who want to see a return on their investment in AI tools sooner rather than later.

At the same time, a steady stream of headlines about job losses due to AI creates a backdrop of uncertainty and pressure even for non-frequent users.

This environment can also increase our tendency to self-criticise. The Change Lab’s research shows employees often using AI show 20 per cent less self-compassion as they struggle with new demands.

AI is often held up as a super-tool with endless capabilities, explains McQuaid – so, when our efforts to use it productively come up short, we tend to blame ourselves rather than the tools.

“The tools themselves are also somewhat unpredictable. And that makes self-criticism a really easy thing for us to reach for,” she says.

“Sometimes the AI tool creators update or make changes to models without even announcing it, so suddenly what worked yesterday doesn’t seem to be working today. And we have no control over that.”

The lack of visibility around the inner workings of the AI model means users are sometimes denied the satisfaction and the dopamine reward that comes from diagnosing a problem and fixing it themselves.

As people lean on AI to draft, summarise or structure their work, they may also begin to doubt the strength of their own judgement or expertise without it.

Wellbeing can also falter if users fall victim to the ‘sunk cost fallacy’ when AI doesn’t give them the output they hoped for. After spending significant time trying to coax the right response from the AI model, users often realise they could have completed the task faster on their own. 

Instead of saving effort, the technology ends up increasing frustration, time pressure and cognitive load.

“There aren’t clear manuals or procedures [for AI] – things are changing all the time. So how we support each other’s learning and self-compassion in the process as we go is really important.” – Dr Michelle McQuaid, founder of The Change Lab

How should HR respond?

Recognising these pitfalls is critical for leaders trying to introduce AI responsibly. Without an understanding of how these tools affect wellbeing, confidence and cognitive load, well-intentioned productivity initiatives could end up creating new psychosocial risks.

McQuaid recommends conducting an “AI human impact assessment” before deploying any new AI tools. This involves assessing technical requirements alongside psychosocial impacts, by asking questions such as: 

• What new cognitive or emotional demands could this introduce?
• What human skills might atrophy?
• How will this change daily interactions?
• Where might connection be lost?

It’s also important that employees themselves are educated on the psychological traps that come with AI, so they can put their own personal guardrails in place.

“[For example], I have a rule for myself where I set an alarm for 60 minutes if I’m working with AI tools, and then I step away, because I know at the end of that 60 minutes, my own clarity and energy start to fade,” says McQuaid.

Gain the skills to navigate the complex landscape of psychosocial wellbeing in the workplace with AHRI’s Psychosocial Code of Practice short course.

How can leaders prevent AI from eroding workplace civility?

As employees become more accustomed to working with AI, questions are emerging about how these interactions might influence civility and relationships in the workplace.

Given that courtesy is not necessary in interactions with AI, many users end up stripping communication back to terse instructions, explains McQuaid. They write things like “Make it better!”, or “Do it again, but shorter.”

“If we’re spending quite a bit of time on these tools, and we’re very direct and blunt when it’s not doing what we need it to, that [could] shift our neural wiring [and impact] how we communicate and engage in our real world,” she says.

This is backed up by The Change Lab’s research, which found that workers often using AI report 19 per cent lower civility scores (67) versus those rarely using AI (86).

As AI becomes more embedded into everyday work, the risk of relationships eroding only intensifies.

“Our relationships are one of our most important [assets] at work, because it’s in our relationships with each other that we expand our resources,” says McQuaid.

“While AI tools do have access to some resources that we don’t, their ability to think outside of a predictable model is quite different from the creativity I see happen between two human beings, who bring together different perspectives and lived experiences, and the energy they get when they bounce off each other.”

How should HR respond?

Managing AI’s impact on human connection is particularly challenging for employers, because any erosion of civility will likely happen gradually and largely out of sight. 

One step leaders can take is to establish AI peer support networks, connecting workers at different stages of AI adoption. Those who’ve navigated the struggle phase can share knowledge and normalise the experience for newcomers, enhancing connection and empathy.

“There aren’t clear manuals or procedures [for AI] – things are changing all the time. So how we support each other’s learning and self-compassion in the process as we go is really important. 

“It also undoes some of the masking that goes on when people think it’s just them that can’t make a tool work. People sometimes [assume] that everyone else has figured it out. But they haven’t, so we need to be more transparent about that.”

For executives and boards focused on productivity and performance, the longer-term human impacts of AI can be easy to overlook. HR is often the first to spot those signals – and, by bringing a psychosocial safety lens to executive discussions, HR can ensure AI supports performance without eroding the conditions that make it possible.

AHRI’s short course on Embedding Responsible AI: Shaping Workforce Planning and Culture will give you a tailored, actionable roadmap to embed HR-specific AI governance, create future-focused workforce plans and a culture that responsibly embraces AI.