AI Doesn't Reduce Work. It Intensifies It.

I need to be honest with you about something. AI did not make my work easier. It made my work the most intense it has ever been.

On any given day, I am running up to ten parallel sessions across Cowork, Claude Code, and various AI chats. I am reading and synthesising more documentation than I ever thought possible. I am managing AI agents the way a floor trader manages open positions, constantly scanning, adjusting, redirecting. My typing has become a bottleneck, so I stopped. I now dictate almost everything through Wispr Flow at nearly 180 words per minute. I built a custom Swift app that reads documents and emails aloud to me so I can listen to one thing while writing another. I have a problem now and my first instinct is not to Google it. It is to code my way out of it.

I have not searched Google in weeks.

And based on the research that has landed in just the past few weeks, I am far from alone.

---

What the Research Says

The Berkeley Study: Task Expansion is Relentless

An eight-month UC Berkeley study published in Harvard Business Review tracked about 200 workers at a U.S. tech company and found something that will resonate with anyone deep in AI tools right now.

Workers did not slow down when AI made tasks faster. They expanded. Designers started coding. Researchers started engineering. People “just tried things” with AI until those experiments quietly became permanent additions to their job descriptions. The researchers call it task expansion, and it is relentless.

The conversational interface of AI tools made work feel like chatting, which dissolved the natural boundaries between work and not-work. People were prompting over lunch, during meetings, in the margins of their evenings.

The BCG Study: AI Brain Fry

Then came the Boston Consulting Group study that gave this experience a name: AI brain fry. They surveyed nearly 1,500 full-time workers and found that 14% of AI users are experiencing genuine cognitive overload from AI tool management.

The numbers are sharp:

And here is the kicker that should keep every CHRO awake at night:

The ActivTrak Data: Where the Time Goes

The data from ActivTrak, tracking over 10,000 workers, fills in the rest of the picture.

AI did not free people to think bigger. It filled every freed minute with more coordination, more communication, more context-switching.

The Productivity Paradox

So the productivity paradox is real. Individual task-level gains of 14 to 55% are well documented across multiple studies. But 89% of managers report zero change in organisational productivity over three years. The speed goes somewhere, but it does not go to the bottom line. It goes to more work.

---

What I am Seeing on My Own Team

I am seeing this play out in real time. People are doing things far outside their roles. Our designers are shipping code. Our strategists are building prototypes. A colleague told me the other day, point blank, that they were in existential crisis mode. They were questioning the value of their work because AI can now do it.

My answer was this: you have a skill set that allows you to get extraordinary work out of these tools. That is not nothing. That is everything.

I am still getting copy-and-pasted ChatGPT emails from clients who clearly never read the output and have no understanding of what it says. A skilled developer using AI can build a website that is worlds apart from what someone with no context can paste together.

---

Pushing Back on the Narrative

Here is where I want to push back on the narrative, though. Because the framing of most of these articles is cautionary. Slow down. Set boundaries. Practice “intentional pauses.” And that is fine advice. But it misses something important about what is actually happening to the people who have pushed through the intensity curve and come out the other side.

When I stopped typing and started dictating, something changed. When I built a tool to read things aloud to me, something changed again. When I learned to manage ten AI sessions like a portfolio rather than a to-do list, the nature of my work shifted fundamentally.

The value I create is no longer in the individual output. It is in the connections between outputs, the pattern recognition across parallel workstreams, the judgment calls about which AI session to trust and which to redirect.

This is the part nobody is writing about yet. Yes, AI intensifies work. But for some of us, that intensity is not a bug to be managed. It is the new shape of knowledge work itself. The question is not how to turn down the volume. The question is how to build the cognitive infrastructure to operate at this tempo without burning out.

---

The Survival Framework

I think the answer involves a few things the research hints at but does not fully articulate.

1. Modality Shifting

I do not read and write the way I did a year ago. I listen and speak. Moving information between senses is not a hack. It is a genuine cognitive relief valve that prevents the single-channel overload the BCG study describes.

2. Agent Management as a Skill

The Berkeley study found that productivity gains peak at three simultaneous AI tools and then decline. But that threshold is not fixed. It is a function of how well you have developed your ability to context-switch between agents. Like any skill, it can be trained.

3. Deliberately Unstructured Time

Not the “intentional pauses” the HBR article recommends, which still feel like managed productivity. I mean genuinely unstructured time where you are not optimising anything. A walk with no podcast. A meal with no screen. The intensity makes this non-negotiable rather than optional.

---

The Uncomfortable Truth

AI did exactly what it promised. It made us more capable. It just turns out that the result of being more capable is not relaxation. It is intensity.

If you are feeling this, you are not behind. You are not doing it wrong. You are in the middle of a genuine phase change in how knowledge work operates.

I would love to hear: how has your daily workflow physically changed since you started using AI tools heavily? Are you typing less? Talking more? Managing more parallel threads? Or have you pulled back? There is no wrong answer. But I think this conversation matters more than another benchmark comparison.