AI Adoption & Productivity
Why Your AI Tools Are Making You Busier, Not Better
An Agile Framework for Sustainable AI Adoption
If your AI workflow looks like 12 browser tabs and three chat windows, you're not innovating. You're context-switching yourself into exhaustion.
I know because I've been there. Recently.
It was the end of a ten-hour day building my consulting practice. I'd been researching market positioning in Claude, pressure-testing frameworks in Gemini, exploring competitive landscapes in Kimi. I was prompting, iterating, comparing outputs, opening new threads when old ones got unwieldy. The energy of working alongside AI is seductive. Every response feels like momentum. Every new chat window feels like a door opening.
When I finally pushed back from my desk that evening, mentally spent and eyes burning, I asked myself one question: What did I actually finish today?
The honest answer? I couldn't point to a single completed outcome. I'd done an enormous amount of work: hours of research, dozens of conversations across multiple tools. None of it had crossed a finish line. I'd been everywhere and nowhere at the same time.
And this wasn't the first time. In my previous role leading learning strategy for one of the largest private employers in the world, I'd felt the same pull. Exploring AI was exciting, and it should have been part of my work. Some days, the exploration became the job. The tools were so responsive, so eager to keep going, that I never hit a natural stopping point. There was always one more prompt, one more angle, one more tool that might give me a better answer.
It was a deceptive trap. The feeling of productivity without the evidence of progress.
The Research That Named What I Was Feeling
Then earlier this year, UC-Berkeley researchers Aruna Ranganathan and Xingqi Maggie Ye gave language to what I'd been experiencing. Their eight-month ethnographic study, published in Harvard Business Review, followed knowledge workers through their daily AI-assisted workflows and reached a clear conclusion: AI doesn't reduce work. It intensifies it.
Not because the tools don't work. They work too well, and in doing so, they reshape the nature of work itself through three mechanisms I recognized immediately.
Task Expansion
AI fills knowledge gaps so effectively that workers begin absorbing responsibilities that were never theirs. The researcher starts handling engineering tasks. The product manager writes code. The senior leader becomes what Ranganathan and Ye describe as a quality-control inspector for an unreliable but prolific junior colleague, spending hours reviewing, correcting, and validating what AI produces. The work didn't disappear. It multiplied.
Blurred Boundaries
AI eliminates what the researchers call “blank page paralysis.” When starting a task requires nothing more than typing a conversational prompt, there's no friction to prevent work from bleeding into lunch, into evenings, into weekends. That “quick last prompt” before stepping away becomes the norm. The breaks that once provided cognitive recovery stop happening.
Cognitive Overload
Workers managing parallel workflows, tracking their manual process alongside AI-generated alternatives, running multiple agents simultaneously, and reviving long-deferred tasks because AI could “handle them,” reported what the researchers captured perfectly:
A sense of always juggling, even as the work felt productive.
That was my ten-hour day.
These three forces don't operate independently. They form a self-reinforcing cycle. AI accelerates tasks, which raises expectations for speed, which drives workers to lean harder on AI, which widens scope, which increases work density. The cycle repeats with no natural stopping point.
I lived that cycle. In my Fortune 1 role, I went from exploring AI tools for our learning strategy to running learning analytics reports on a Saturday, squeezed between family events, because the AI made it possible, which somehow made it feel necessary. No one asked me to work that Saturday. The tool just made it easy enough that I couldn't justify not doing it.
The Hidden Cost
The individual experience scales into an organizational crisis faster than most leaders realize.
Upwork Research Institute, 2024
DHR Global Workforce Trends, 2026
MIT Project NANDA
Burnout is concentrated at associate and entry-level ranks, the very people organizations are asking to adopt AI fastest. Among C-suite leaders, 62% fewer report that burnout is affecting their engagement at work. The people making AI adoption decisions aren't operating under the same conditions as the people living with those decisions.
The tools are making people busier. Burnout is rising. Organizations aren't seeing returns. That's not a technology problem. It's a process problem, and it's one I've seen before.
The Self-Reinforcing Trap
Jevons Paradox
Economists have a name for what's happening: the phenomenon where efficiency gains in using a resource cause total consumption of that resource to rise, not fall. We've seen this pattern before. AI is following the same trajectory.
Email promised to replace memos and streamline communication. Instead, it created an always-on culture where the average professional now manages 120+ messages a day. Spreadsheets didn't eliminate analysts. They enabled infinitely more complex models that required more analysts to build and maintain. Self-checkout didn't reduce staffing needs. It shifted cashiers into oversight roles while adding new layers of exception handling.
The tool gets more capable, so we give it more to do. It handles more, so we expect more. We expect more, so we work more. We call it innovation.
Without some kind of structural intervention, this cycle has no natural braking mechanism. Individual discipline isn't enough. Telling people to “set boundaries with AI” is the 2026 equivalent of telling people to “manage their inbox better” in 2005. It didn't work then. It won't work now.
What does work is process. And there's a proven methodology that was built specifically to manage complexity, limit work in progress, and create intentional stopping points in fast-moving environments.
It's called Agile. And it's not just for software teams anymore.
The Agile Connection
Before I founded B.Human Solutions, I led a large-scale Agile transformation for a project management organization inside the merchandising arm of a Fortune 1 company. The results: a 35% increase in team capacity and over 8,000 hours reclaimed annually. The insight that made it work wasn't a tool or a framework diagram. It was a mindset shift.
The Three Principles That Drove the Transformation
1. Planning before motion.
We invested time upfront in understanding the challenge before anyone started building. It felt slow at first. After their first projects completed, something shifted: less rework, fewer surprises, far fewer moments where the team scrapped weeks of effort because they'd built the wrong thing fast.
2. Ownership closest to the task.
Project managers stopped absorbing work that belonged to other team members. When accountability sat with the person doing the work, busy work evaporated and decisions happened faster.
3. One activity at a time.
Managed visually through kanban boards so every team member could see what was in progress, what was waiting, and what was done. No more invisible multitasking.
Those three principles map directly onto AI work.
Planning before motion. Most people open an AI tool without a clear definition of what they're trying to accomplish or what “done” looks like. They start prompting and see where it goes.
Ownership closest to the task. AI makes work possible that isn't actually yours to do. The Saturday analytics report no one asked for. The competitive analysis outside your scope. The rabbit hole that felt productive but served no defined objective.
One activity at a time. The average knowledge worker running AI-assisted work has multiple conversations open in parallel, several tabs, and half-finished threads they intend to “come back to later” but probably won't.
The researchers at UC-Berkeley proposed what they called an “AI Practice” framework: intentional pauses, sequenced work phases, and protected time for human reflection. If you've worked in Agile, you recognize those recommendations immediately. They're sprint retrospectives. They're backlog prioritization. They're work-in-progress limits. They're daily standups.
The principles aren't new. They're just newly urgent.
The Agile AI Sprint
Here's a structure I've started applying to my own AI work. It's not a complete methodology. It's a starting point, and one that's already changed how I work.
The Agile AI Sprint
Five disciplines for sustainable AI productivity
Plan the sprint before opening a single tool
Take five minutes. Write what specific outcome you need to produce. Define what "done" looks like. If you can't articulate it in one sentence, you're not ready to start prompting.
One tool, one task, completion
Work in a single AI environment on a single objective until it's finished or you've consciously decided to stop. Capture new ideas in your backlog—don't open a new tab.
Maintain a backlog
Every idea, tangent, and "I should explore this" thought goes into a simple list, not into a new browser tab. The backlog makes your invisible work visible and gives you control over when to pursue it.
Run a daily standup with yourself
End each session with two minutes answering three questions: What did I finish? Where did I get stuck? What's the single most important thing for my next session?
Know when to stop
If you defined your outcome at the start, you'll know when you've reached it. Close the tabs. Step away. The tools will be there tomorrow, and you'll come back sharper for having left.
The Business Case
Let's do the math that most organizations are avoiding.
Research by UC Irvine's Gloria Mark found that after an interruption, workers take an average of 23 minutes before returning to their original task. If your team members are switching between AI tools, threads, and manual workflows ten times a day (a conservative estimate), that's nearly four hours of lost productive time per person, per day. Not lost to laziness. Lost to the structural absence of a process that manages focus.
Multiply that across a team. A department. An enterprise. Then layer in the human cost.
The tools aren't broken. The approach is.
Where to Start: Tomorrow, Not Next Quarter
Before your first AI interaction of the day, write one sentence defining what you need to produce and what done looks like. Work in one tool on that single objective until it's complete. Capture every tangent in a backlog instead of a new tab. End the day with two minutes answering: what did I finish, where did I get stuck, what's the one thing that matters most tomorrow?
Make AI workflow visible. You wouldn't let a project team run five parallel workstreams with no kanban board and no sprint plan. Create shared norms around tool selection, task focus, and structured reflection. Run a weekly retrospective that includes AI-assisted work, asking not just what they're using AI for, but whether it's producing completed outcomes.
Stop measuring AI adoption by usage rates and tool deployments. Start measuring it by completed outcomes, rework reduction, and whether your people are burning out in the process. When only 5% of enterprise AI pilots reach production with measurable impact, the problem isn't the model you chose. You deployed powerful technology into an environment with no process to harness it. Agile solved this for software development two decades ago. It can solve it here.
Your people don't need more AI tools. They need a structure that makes the tools they have actually work, for the business and for them.
The self-reinforcing cycle the researchers describe has no natural braking mechanism. Efficiency creates demand. Demand creates overload. Overload creates burnout. And burnout doesn't show up in your AI adoption dashboard.
The methodology exists. The evidence is there. The decision is whether to act on it before the cycle runs long enough to cost you people, capacity, and trust.
Build a Sustainable AI Practice
B.Human Solutions partners with executive leaders to design AI adoption strategies that deliver results without burning out your people.