The AI Productivity Mirage: More Email, Less Deep Work

AI was supposed to shave hours off our calendars. It’s done the opposite—at least for now. Fortune reports that time spent emailing has doubled and focused work sessions have fallen by 9%. Those two numbers shouldn’t sit beside each other quietly; they should set off alarms.

Here’s the uncomfortable part: both of those trends can be true even in a company that thinks it’s “winning” with AI.

Who benefits when every meeting spawns three follow-ups and every task spawns an AI-generated draft that someone else then edits? Follow the money. Vendors make more when you prod, prompt, and publish more. Internal champions get credit for “AI adoption” while teams inherit a new kind of administrative cascade—short messages, rapid revisions, extra coordination. Convenient, isn't it.

I’ve seen this movie before. When email first hit offices at scale, executives promised less paperwork and faster decisions. What we got was inbox addiction and a new layer of soft obligations—“quick thoughts?”, “circling back”, “just bumping this.” AI is replaying that pattern at high speed. The tech accelerates production; humans absorb the overhead.

The Fortune piece nails the symptom: more time in email, fewer long, uninterrupted blocks of concentration. Those focused sessions—where analysis, synthesis, and creative problem-solving happen—are the scarce currency of knowledge work. A 9% drop matters. It’s not a comfort metric; it’s the difference between solving a knotty problem and rustling through a pile of messages all afternoon.

But shorter focus doesn’t automatically mean smarter work, and that’s where the current narrative gets slippery.

AI is being treated like a tool you drop into an existing workflow; that’s the mistake. You don’t bolt a turbocharger onto a car and call it a train. Generative models excel at producing text fast—emails, summaries, bullets. Speed creates volume. Send more memos and you get more responses. Produce polished drafts and someone has to review, localize, tailor, and align them with policy. The net effect can be more friction, not less.

Tech without re-engineering is just labor displacement dressed up in launch-deck gloss. Companies often buy AI as if software alone redesigns work. It doesn’t. You have to redraw responsibilities, change approval chains, set clear ownership, and redesign information flows to account for machine-generated content. Otherwise the machine just multiplies handoffs. Ask any operations manager: more intermediaries mean more delay.

Here’s what they won't tell you: many vendor metrics measure API calls or seats sold—not worker well-being or decision quality. That misalignment shapes corporate incentives. You end up with tool stacks optimized to increase usage and justify licensing fees, not to reduce cognitive load. That’s not a conspiracy; it’s a predictable market dynamic.

Look at how CRM platforms spread. Salesforce didn’t conquer boardrooms by proving salespeople were happier; it won by showing executives more data and more “activity.” Now replace manual entries with AI-generated notes and summaries. Unless someone rethinks quotas, review cycles, and how decisions are actually made, you get the same pattern: more logs, more dashboards, less thinking time.

The Fortune headline speaks broadly, but “employees” aren’t a monolith. Roles heavy on routine text—customer support, recruiting, legal redlining—may see different trajectories than roles demanding deep conceptual thought—product strategy, R&D, creative work. If AI accelerates first-order tasks in support roles, email volumes might still rise because downstream teams need context and decisions.

We also need to be honest about what “focused work sessions” are measuring. Does starting a timer signal deep engagement, or does it punish necessary switching between documents, tools, and people? Are short bursts of high-intensity coordination being misclassified as distraction? These definitional choices matter because they drive managerial response—ban the tools, over-police app time, or actually redesign the work.

Some will counter: give it time; AI will streamline processes, automate repetitive work, and make us more productive long-term. That’s plausible. Every major office technology—from spreadsheets to group chat—had a messy adoption curve. But that argument quietly assumes organizations will do the heavy lifting: training staff, redesigning processes, changing performance metrics from “activity” to “outcomes.”

Many won’t. They’ll celebrate pilot programs and vendor case studies while riding a temporary spike in output metrics that masks declining deep work. The dashboard looks green. The humans feel red.

Here’s a simple operational test, and no consultant is required: pick a recurring task and map every handoff that AI touches. Who drafts, who edits, who approves, who re-phrases for yet another system? If the count goes up, you’ve introduced friction. If it falls, you’ve redesigned successfully. That’s trivial to measure—and most firms aren’t doing it.

Follow the money again. The version of AI being sold as a productivity booster is often optimized for the vendor’s revenue model, not for human cognition. That’s not something a CEO quote will fix. It’s something operations teams and labor leaders will have to force into their planning—and fast.

If doubled email time and a 9% decline in focused sessions become the new baseline, the real story won’t be about “AI adoption rates”; it will be about how many promising ideas quietly die in inboxes no model was ever trained to see.