Almost every UC vendor pitch in 2025 and 2026 has led with the same promise: AI will save time, automate tasks, and free employees to focus on higher-value work. Meeting summaries, message drafts, automated workflows: the productivity gains sound inevitable.
But a new eight-month study from UC Berkeley researchers, published in Harvard Business Review, has found the opposite. Tracking roughly 200 employees at a US tech firm, professors Aruna Ranganathan and Xingqi Maggie Ye discovered an AI productivity paradox: finding that AI tools didn’t reduce work but consistently intensified it:
“On their own initiative workers did more because AI made ‘doing more’ feel possible, accessible, and in many cases intrinsically rewarding.”
Employees voluntarily worked at a faster pace, took on a broader scope of tasks, and extended work into more hours of the day. One engineer captured the reality: “You had thought that maybe, oh, because you could be more productive with AI, then you save some time, you can work less. But then really, you don’t work less. You just work the same amount or even more.”
The question for UC leaders is simple, then. Is this already playing out inside your collaboration stack?
The Data Says It’s Already Happening
Independent research from Microsoft, PwC, and Gartner confirms the same pattern at scale — and specifically inside the UC platforms enterprises are deploying today.
Microsoft’s Work Trend Index, released in June 2025 and based on trillions of Microsoft 365 productivity signals, revealed what it calls the “infinite workday.”
It found the average employee now receives 117 emails and 153 Teams messages daily. Workers are interrupted every two minutes: equivalent to 275 times per day.
Key findings on AI work intensification:
- 40% check email before 6am
- 29% are back in their inbox by 10pm
- Evening meetings after 8pm, up 16% year-over-year
- 48% of employees say their work feels chaotic and fragmented
PwC’s 2025 Global Workforce Hopes & Fears Survey, meanwhile, which surveyed nearly 50,000 workers across 48 countries, found that 35% of the global workforce feels overwhelmed at least once a week — rising to 42% among Gen Z. Yet only 14% of workers are using generative AI daily. Ironically, those daily users report significant gains in productivity (92%), job security (58%), and salary (52%).
According to Pete Brown, Global Workforce Leader at PwC:
“Employees using AI every day are reaping the rewards – higher productivity, greater job security and better pay. But to scale these benefits, businesses must go beyond training. Work itself needs to be redesigned.”
Gartner’s 2026 Future of Work Trends for CHROs lists “AI’s biggest hidden cost: your employees’ mental fitness” as a top-nine trend. The research firm also flags “AI workslop” — low-quality AI output that employees spend hours reviewing and fixing — as organizations’ top productivity drain.
Three Ways It Shows Up in UC
The Berkeley study identified three forms of work intensification. Each maps directly onto dynamics inside unified communications platforms.
Role Creep via AI Copilot
Because AI copilot tools fill gaps in knowledge, workers increasingly step into responsibilities that previously belonged to others. The Berkeley researchers found product managers writing code and researchers taking on engineering tasks.
When collaboration platforms make it easy for anyone to draft workflows, write scripts, or automate processes, role boundaries dissolve. The unintended consequence? Specialists — particularly engineers — get pulled into reviewing, correcting, and guiding AI-generated or AI-assisted work produced by colleagues.
The Always-On Collaboration Layer
Workers in the study sent “one last prompt” before leaving their desk, used AI during lunch breaks, and prompted tools during meetings. The conversational interface made starting a task feel closer to chatting than undertaking formal work.
In Microsoft Teams, Slack, or Webex, the line between messaging a colleague and prompting an AI assistant has collapsed entirely. This is precisely why 29% of workers are back in their inbox by 10pm: the friction between work and non-work has been removed, and the collaboration platform is where it happens.
Agent Sprawl in UC Platforms
As unified communications platforms push agentic AI tools — Copilot Studio, Zoom AI Companion, Salesforce Agentforce — workers manage multiple autonomous threads simultaneously. The Berkeley study found this created “a continual switching of attention, frequent checking of AI outputs, and a growing number of open tasks.”
While workers described having an AI “partner” that helped them move through their workload, the reality was constant context-switching and cognitive load. The more agents you deploy, the more human oversight they require — a new management layer nobody budgeted for.
What UC Leaders Should Do About the AI Productivity Paradox
The Berkeley researchers propose that organizations develop an “AI practice” — intentional norms and standards around AI use. For UC leaders addressing the AI productivity paradox, that translates into four concrete actions:
Enforce Focus Time at the Platform Level
Do-not-disturb scheduling, notification batching, and protected focus windows already exist in Teams, Zoom, and Slack. But these features only work if they’re policy. Configure them organization-wide and make disconnection a cultural norm
Govern AI Agents Before They Sprawl
As AI copilot and agent capabilities expand across your UC stack, establish governance now: who can deploy agents, what they can act on, and how outputs get reviewed.
Audit AI-to-Human vs. Human-to-Human Time
The Berkeley researchers emphasize “human grounding” — short opportunities to connect with others that interrupt continuous solo engagement with AI tools and restore perspective. Protect 1:1s, team retrospectives, and unstructured collaboration time.
Measure Outcomes, Not Just Speed
“Hours saved” is the vendor metric. Track decision quality, error rates, rework cycles, and employee wellbeing alongside throughput. The Berkeley study warns that what looks like higher productivity in the short run can mask silent workload creep and growing cognitive strain.
