What workers really think about workplace AI assistants

Written by: Talitakuum Ekandjo, Lecturer, Te Herenga Waka — Victoria University of Wellington

Imagine starting your workday with an AI assistant that not only helps you write emails ^[1] but also tracks your productivity ^[2], suggests breathing exercises ^[3], monitors your mood and stress levels ^[4] and summarises meetings ^[5].

This is not a futuristic scenario. Workplaces globally are already quietly transforming into AI-powered environments, with 75% of knowledge workers using AI tools such as Microsoft Copilot, OpenAI’s ChatGPT and Microsoft Viva Insights, according to recent Microsoft data ^[6].

Most (85%) find these tools effective at prioritising tasks. A growing “bring your own AI” trend has also seen 78% of employees introducing their own AI solutions to find relief from endless emails, meeting overloads and blurred work-life boundaries.

As part of my research ^[7], I explored how the use of AI assistants works out in practice and how it is transforming our work experiences.

The study is based on interviews and self-reflection journals kept by workers across government, technology, education and finance sectors in New Zealand. My focus is on Microsoft Viva Insights, a productivity assistant which leverages AI to analyse work habits, offer personalised recommendations to optimise workflows and encourage healthier routines in the workplace.

My findings highlight benefits but also reveal a stark contrast between AI’s promises and workplace realities.

A rear view of a woman stretching while seated at her desk.

Office AI assistants can encourage healthier routines in the workplace. Getty Images ^[8]

Potential for productivity and wellbeing

For many workers, the AI assistant offered tangible benefits. They discovered it could guide them into setting goals and prioritising tasks. As a result, they felt they could allocate their time and efforts more efficiently and effectively, which transformed how they tackle workloads.

The AI assistant functioned as a self-monitoring and reflection tool. Workers described it as a “dashboard”, “safety net” and “tracker” that helped them notice work patterns they might otherwise overlook. For example, by consolidating key meeting documents, the AI assistant helped them stay organised, prepared and efficient.

Beyond improved productivity, workers also discovered unexpected personal benefits. Some found that following the breathing exercise suggestions created a “settling feeling” and served as a “good reminder” to prioritise self care and not solely focus on work. Workers also described the AI assistant’s potential to facilitate interaction among “introverted” or “less socially active” team members.

However, the findings also revealed significant limitations.

The AI rigidity trap

Workers frequently found the AI assistant oversimplified the messy, interconnected and fluid nature of modern work, especially because it does not know about or understand the demands in other parts of workers’ lives.

The AI assistant’s suggestions often proved impractical for roles that demand constant availability and real-time collaboration. John, a software developer whose AI assistant regularly blocks off “focus time”, explained:

I cannot have any notifications off. My job right now […] a lot of it is collaborative and working with other people.

Kyle, an applications support manager, echoed this sentiment:

Booking focus time is lovely in theory, but I never have two hours free on my calendar. I’ve looked at different ways of properly focusing attention, for example shutting down the email, but people end up at your desk. So, focus time isn’t necessarily focus time.

Digital overwhelm and distraction

Workers found interacting with the AI assistant was an additional task to manage and its multiple interaction channels could be overwhelming. The effort and time needed to “interact with or research these tools to try and make them better” often felt counterproductive as it took time away from actual work.

The AI assistant’s personalised nature also raised privacy concerns. Workers indicated they were “always suspicious” because they never quite knew who is reading and analysing their data. The mood-tracking feature, which invites workers to pick an emoji from a menu of expressions that most closely matches their mood, exposes this privacy dilemma.

Workers mentioned they “do the smiley faces” but are “always a bit worried” when selecting frowning faces because they are not entirely sure if anyone is recording and judging them.

AI assistants hold immense potential, but their success depends on how well they align with the complex nature of work.

My research suggests workers are most likely to embrace these tools when they feel a sense of agency and understanding. As AI continues to reshape our workplaces, the key will be creating technology that serves humans and adapts to their work realities.

The most effective AI will not be the most powerful but the most flexible and adaptable to accommodate diverse roles and role-specific work patterns. To foster trust and acceptance, workers need to be given control to determine what data the AI assistant uses when making suggestions in specific work scenarios.

Finally, data transparency is crucial to ensuring workers’ confidence that their privacy is respected, and their data is used ethically.