A few months ago I started more projects in a single quarter than I had in the previous two years combined. Newsletter. Calculator site. Digital products. Podcast content. Agent workflows. All of it felt possible and easily obtainable. AI was handling the first drafts and it was fast and engaging and enjoyable to keep pushing. The capacity felt almost limitless, though it was definitely not without some frustrating sycophantic loops, the kind where the tool agrees with everything you say instead of telling you what is actually wrong. And then I looked at what was actually finished, and at my original goal to increase both productivity and free time. I cannot honestly say that was a clear goal achieved.
That gap between what felt achievable and what shipped is the most honest thing I can tell you about AI. It does not give you more time. It gives you more surface area. What you do with that surface area is now the most consequential professional decision most of us are making, and almost nobody is making it deliberately.
The Data Observation: The Number That Should Change How You Read Your Calendar
In February 2026, researchers from UC Berkeley published an eight-month study of 200 employees at a U.S. technology company. The finding contradicted almost every productivity promise attached to AI adoption. Workers using AI tools did not work less. They worked more. The researchers described a self-reinforcing cycle: AI accelerated certain tasks, which raised expectations, which pulled more work into the frame, which AI then helped accelerate again. The cycle did not produce more rest. It produced more open tabs.
This is not an argument against using AI. It is a description of what happens when you use it without a framework for what the recovered time is actually for.
The same dynamic shows up in the labor market data, just at a larger scale. A Stanford study led by Erik Brynjolfsson, using payroll data from millions of workers through late 2025, found that entry-level employment in highly AI-exposed occupations like software engineering and customer service declined by 13% since late 2022, while employment for older, more experienced workers in the same jobs grew. The pattern is specific. AI is not hitting the workforce evenly. It is compressing the bottom of the task pyramid first, the repeatable, codified, first-pass work that used to fill junior roles and, frankly, large portions of many senior ones too.
The hours those tasks used to occupy are now available. The question of what fills them is not a technology question. It is a judgment question.
The Decision Matrix: What You Actually Do With the Time AI Returns
There are four ways people fill the hours that AI reclaims, and most people cycle through all four without realizing they are making a choice at all.
The first is depth. You use the recovered time to go further into the work that actually requires you: relationships, strategy, judgment calls that need context AI does not have. This is the version every productivity article promises. It is also the least common default behavior, because depth requires deliberate choice and resists the pull of momentum.
There is another layer to this that most people do not talk about. AI gives you speed before it gives you depth. In the early stages of any AI workflow, the first pass comes fast and feels complete. Getting genuinely deep output requires familiarity with the tool's structure and limitations, knowing where to push back, where to add context, and where the confident-sounding answer is actually surface level. That fluency takes time to build, which means the depth option is harder than it looks even when you have chosen it intentionally.
The second is volume. You produce more of the same category of work at the same level. More decks, more reports, more content. Output increases. The ceiling stays where it was. Volume is the easiest default because it fits the existing definition of what good looks like.
The third is expansion. You start new projects because new projects feel achievable in a way they never did before. This is where I spent most of last quarter. The trap here is not laziness. It is the opposite. Everything genuinely looks possible when AI is handling the scaffolding. The shiny object problem in an AI-augmented workflow is not a motivation failure. It is a prioritization failure wearing the mask of productivity. Projects accumulate. Attention divides. Nothing ships at the level it deserved.
The fourth is recovery. You use the time to rest, reconnect, and restore the judgment capacity that makes the first option possible. The Berkeley Haas researchers who documented AI work intensification proposed exactly this as the antidote: intentional pauses built into the workflow, not as a productivity hack, but as a structural defense against the quiet accumulation of overload. Almost nobody chooses recovery on purpose. The organizations that build it into the system tend to produce the most durable results.
Most people reading this are somewhere between the second and third options. The question worth sitting with is not whether you are using AI. It is what you have decided the recovered time is in service of.
The issues that matter most to your career and your team do not get easier because you have more bandwidth. They get more exposed. If your highest-value work is relationships, influence, judgment, and the kind of tacit expertise that comes from years in the arena, then the hours AI returns are most valuable when pointed directly at those things. Not at the next project that looks shiny and achievable.
And there is one more option that does not fit neatly into any productivity framework. You could use the time to show up more fully for the people in your life who matter. AI should make our lives better, not just more efficient. Using it to create space for real human connection rather than more output might be the best possible return on the investment.
The Short Version: AI does not give you more time. It gives you more surface area, and most people fill that surface area without ever deciding what it is actually for.
Try This Now: Pick one project you started in the last ninety days that involved AI in any meaningful way. Write one word next to it: depth, volume, expansion, or recovery. Be honest about which category it actually falls into, not which one you intended. If the answer surprises you, that is your signal. The category you keep defaulting to is the one worth examining before you start the next thing.
Carlo DelDonno, MD | Army Captain — The Leadership Growth
If this was useful, forward it to one person. That's the whole ask.
Pro Tier: The Leadership Conversation
### Navigating Output, Expectations, and the End of the Time-for-Wages Contract
For most of modern work history, the wage agreement has been a proxy. You paid someone for their time because time was the closest measurable stand-in for their effort and their output. The correlation was imperfect but stable enough to build compensation systems, performance reviews, and promotion criteria around.
AI is breaking that correlation quietly and faster than most organizations are prepared to acknowledge. A developer who used to spend three days building a feature scaffold can now produce it in four hours. A writer who spent a full day on a first draft can have a working version by lunch. The output is the same. The time is not. And most hiring pipelines, promotion timelines, and workload distribution models were built for the old ratio.
This creates three specific conversations that leaders will need to navigate.
Conversation One: What do we do with the recovered capacity?
The default answer is to add more work. The Berkeley Haas study found that this is exactly what happens by default, and that it leads directly to burnout, decision fatigue, and eroded output quality over time. The question that changes the frame is not "what can you take on now that you have more bandwidth?" It is: what work has been getting less than it deserves because we never had the time?
Conversation Two: How do we measure output when time is no longer the proxy?
Most performance systems measure activity: hours logged, tasks completed, meetings attended. When AI compresses the activity side without changing the output side, those metrics start measuring the wrong thing. Consider a military training environment where every instructor follows the same script. The senior officer still has to rank performance. So what does she evaluate? Presence, judgment under pressure, how someone handles the moment the script breaks down. She is already doing outcome measurement. AI is about to put every knowledge worker in that same position.
A practical starting point: instead of asking what someone is working on, ask what decision they made this week that they could not have made a year ago.
Conversation Three: What happens when someone finishes their work in half the expected time?
If an employee completes their deliverables in twenty hours instead of forty, do you pay them for forty? Do you add more work? Do you let them have the time? EY's 2025 AI Pulse Survey found that organizations are most often reinvesting AI productivity gains into growth, upskilling, and resilience rather than reducing headcount. There is no universal right answer, but the answer needs to be explicit, consistent, and set before the situation creates resentment.
The Short Version (Pro): AI does not give you more time. It gives you more surface area, and what you point that surface area at is now the most consequential professional decision most of us are making. For leaders, that same problem has a harder edge: the old rules about time, output, and what you owe each other are changing faster than most organizations are ready to talk about.
Try This Now (Pro): Pick one task you completed this week with significant AI assistance. Estimate how long it would have taken you eighteen months ago. Now account for the actual time you spent: the first pass, the review, the corrections, the judgment calls the AI got wrong. Calculate the real hours saved. Then ask one question: what did you do with them?
Carlo DelDonno, MD | Army Captain — The Leadership Growth
References
1. Ranganathan, A. and Ye, X.M. "AI Doesn't Reduce Work — It Intensifies It." Harvard Business Review, February 9, 2026. 2. Brynjolfsson, E., Chandar, B., and Chen, R. "Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence." Stanford Digital Economy Lab, August 2025. 3. EY US AI Pulse Survey, Wave 4. Ernst & Young LLP, December 9, 2025. 4. Bick, A., Blandin, A., and Deming, D. "The Impact of Generative AI on Work Productivity." Federal Reserve Bank of St. Louis Working Paper 2024-027C, revised February 2025. 5. PwC Global AI Jobs Barometer 2025. PricewaterhouseCoopers, June 2025.