Nearly every UC vendor pitch in 2025 and 2026 makes the same promise: AI will save time, automate tasks, and free up employees to focus on higher-value work. From meeting summaries to drafting messages to automated workflows, productivity improvements seem inevitable.
But a new eight-month study by researchers at the University of California, Berkeley, published in the Harvard Business Review, found just the opposite. After tracking nearly 200 employees at a US technology company, Professors Aruna Ranganathan and Shin-Chi Maggie Yeh discovered an AI productivity paradox. We have found that AI tools consistently enhance jobs, rather than reduce them.
“Employees volunteered to do more because AI enabled them to ‘do more,’ which they felt was accessible and often inherently rewarding.”
Employees voluntarily worked at a faster pace, took on a wider range of tasks, and worked longer hours in a day. One engineer captured the reality. “We thought AI could improve productivity so we could save time and do less work. But in reality, we’re not working less. We’re just doing the same amount of work, or more.”
So, the question for UC’s leadership is simple. Is this already running within your collaboration stack?
data shows it’s already happening
Independent research by Microsoft, PwC, and Gartner confirms the same pattern at scale, particularly within the UC platforms that enterprises currently deploy.
Microsoft’s Work Trend Index, released in June 2025 and based on trillions of Microsoft 365 productivity signals, revealed what the company calls the “infinite workday.”
What does the average employee currently receive? 117 emails and 153 Team Message every day. Worker’s work is interrupted every 2 minutes: Equivalent to 275 times per day.
Key findings on AI work enhancement:
- 40% check email before 6am
- 29% are back in their inbox by 10pm
- Night meetings after 8pm, up 16% year over year
- 48% of employees say their work feels chaotic and fragmented
Meanwhile, PwC’s 2025 Global Workforce Hopes and Anxieties Survey, which surveyed nearly 50,000 workers in 48 countries, found that: 35% The global workforce feels overwhelmed at least once a week, rising to 42% of Gen Z. only 14% of workers use generated AI every day. Ironically, these daily users report significant increases in productivity (92%), job security (58%), and pay (52%).
Pete Brown, Global Workforce Leader at PwC, said:
“Employees who use AI every day are reaping the benefits of increased productivity, greater job security, and better pay. But to scale these benefits, companies need to go beyond training. They need to redesign the jobs themselves.”
Gartner’s 2026 Future of Work Trends for CHROs lists “AI’s Biggest Hidden Cost: Employee Mental Fitness” as a top nine trend. Research firms have also flagged “AI work cropping” — low-quality AI output that employees spend hours reviewing and modifying — as organizational misconduct. Biggest loss of productivity.
Three ways to appear in UC
The Berkeley study identified three forms of labor intensification. Each maps directly to dynamics within the unified communications platform.
Role Creep with AI Co-Pilot
As AI co-pilot tools fill knowledge gaps, employees are increasingly taking on responsibilities that previously belonged to others. Berkeley researchers found that product managers write the code and researchers handle engineering tasks.
Role boundaries disappear when collaboration platforms make it easy for anyone to draft workflows, write scripts, and automate processes. What are the unintended consequences? Professionals, especially engineers, are drawn to reviewing, revising, and coaching AI-generated or AI-assisted work created by their colleagues.
Always-on collaboration layer
Employees in the study sent a “last prompt” before leaving their desk, used AI during their lunch break, and used a prompting tool during meetings. The conversational interface now feels more like chatting than starting a formal task.
Microsoft Teams, Slack, and Webex completely break down the line between sending a message to a coworker and instructing an AI assistant. That’s exactly why 29% of employees are back in their inbox by 10pm. Collaboration platforms help eliminate friction between work and non-work.
Agent sprawl on UC platforms
As unified communications platforms power agent AI tools (Copilot Studio, Zoom AI Companion, Salesforce Agentforce), workers manage multiple autonomous threads simultaneously. The Berkeley study found that this resulted in “continuous switching of attention, frequent checking of AI output, and an increase in backlogs of tasks.”
Employees described an AI “partner” helping them handle the workload, but in reality, there was constant context switching and cognitive load. The more agents you deploy, the more human oversight you will need. So it’s a new layer of management that no one budgeted for.
What University of California leaders should do about the AI productivity paradox
The Berkeley researchers suggest that organizations develop “AI practices,” or intentional norms and standards for the use of AI. For University of California leaders grappling with the AI productivity paradox, that translates into four concrete actions.
Enforce focus time at the platform level
Do Not Disturb schedules, batch notifications, and protected focus windows already exist in Teams, Zoom, and Slack. However, these features only work based on policy. Set across your organization to make disconnection a cultural standard
Manage AI agents before they spread
As AI copilot and agent capabilities expand across your UC stack, establish governance now, including who can deploy agents, what they can interact with, and how output is reviewed.
AI-to-human and human-to-human time audits
The Berkeley researchers focus on “human grounding,” or taking a break from using AI tools alone and regaining perspective due to brief opportunities to connect with others. Protect time for one-on-ones, team reflections, and unstructured collaboration.
Measure not only speed but also results
“Time Saved” is a vendor metric. Track decision quality, error rates, rework cycles, and employee health along with throughput. The Berkeley study warns that what appears to be a short-term increase in productivity can mask a silently increasing workload and increasing cognitive strain.
