Meeting-Free Fridays emerged from a simple observation: if we fill every available hour with meetings, we leave little room for the work those meetings are supposed to support.
Meetings can be useful. Collaboration matters. But useful things still need boundaries.
I think we may need to apply the same thinking to AI.
AI-Free Fridays.
A protected period to think, learn, create, connect, and appreciate what we have accomplished, with our AI tools temporarily set aside.
I use AI extensively. I see how much it can help us build, understand, explore, and communicate. I want to keep discovering what becomes possible with it.
But I also want us to be deliberate about what happens to the time it saves and the human experiences it changes.
Meeting-Free Fridays protect time. AI-Free Fridays could protect thinking, connection, and the space to feel accomplished.
That last part has become especially important to me.
One of the pain points on my team today is finding room to celebrate achievements. Previously, it felt like there was more time to let an accomplishment sink in. We could finish something difficult, appreciate it together, and enjoy the feeling that we had moved forward.
Today, work can feel like a hamster wheel of productivity.
We finish one thing and immediately see the next five things we could do. The distance between an achievement and the next expectation keeps shrinking.
We are getting better at producing. I want to make sure we still have room to experience what we produce.
AI did not invent this problem. Teams were struggling with endless backlogs, rising expectations, and burnout long before generative AI arrived. But AI can accelerate a pace we have already forgotten to question.
And that makes this a leadership question as much as a technology question.
What do we want greater productivity to give us?
The default response to difficulty
When a tool saves a few minutes, we fit it into an existing workflow. When it saves hours, we reorganize the workflow around it. When it becomes available for almost every intellectual task, it starts changing our default response to difficulty.
| The difficulty | Becomes |
|---|---|
| A blank page | A prompt |
| An unfamiliar codebase | A request for an explanation |
| A difficult decision | A generated list of tradeoffs |
| A moment of uncertainty | Another question for the model |
Each choice can be reasonable. Many are excellent uses of AI.
The cumulative effect deserves attention.
We can become accustomed to receiving an answer before forming our own view. We can start reviewing ideas before practicing how to develop them. We can produce explanations that sound clearer than our actual understanding.
The concern is easy to miss because the output may keep improving.
A polished document does not tell us how deeply its author considered the problem. Working code does not tell us whether the person responsible for it understands its failure modes. A persuasive recommendation does not show whether someone examined the assumptions behind it.
The quality of presentation can exceed the quality of understanding.
As AI makes polished output easier to produce, we need deliberate ways to keep developing the understanding underneath it.
Learning requires participation
An AI-free period creates one opportunity to do that.
Write the first draft. Read the source material. Trace the code. Form a hypothesis about the bug. Sketch the architecture. Decide what you think before asking for alternatives.
Stay with the uncertainty long enough to discover what you understand and where the gaps are.
Some of that effort will feel slower. That is part of the exercise.
We already accept this in other areas of life. Practicing an instrument involves playing passages we cannot yet play well. Physical training involves effort we could avoid with a machine. Learning to speak another language involves the awkwardness of finding our own words.
The process matters because participating in it develops a capability.
The same is true of intellectual work.
Trying to explain a system can reveal a gap that reading a fluent explanation allowed us to overlook. Attempting a design can expose a tradeoff we would otherwise have accepted without noticing. Writing a difficult paragraph can force us to decide what we actually believe.
That does not make every struggle valuable. Repetitive administration, avoidable confusion, and hours lost to poor tooling do not become worthwhile simply because they are difficult.
The useful question is more specific:
Which parts of this effort are developing something I want to keep?
| For | The effort worth keeping might be |
|---|---|
| An engineer | The ability to reason about a system under failure |
| A manager | The ability to give thoughtful feedback |
| A writer | Finding a voice |
| Someone learning a new field | Building the foundation needed to ask better questions |
AI can help with all of these. It can be a patient tutor, a useful critic, and a way to explore unfamiliar territory.
But learning still requires participation.
If every unfamiliar task is immediately delegated, we need to ask where the next generation of expertise will come from. An experienced engineer may recognize a subtle mistake because they have spent years building, breaking, and repairing systems. Someone entering the profession needs opportunities to develop that judgment too.
Faster delivery does not automatically create those opportunities. Teams have to make room for them.
An AI-free period can give people permission to be beginners again: to attempt something, get stuck, ask a colleague, revise their approach, and eventually understand why it works.
That permission matters in a workplace where an impressive answer is always seconds away.
In engineering
In engineering, I would use that time to practice understanding at the level where we remain responsible.
Can I explain how a request moves through this system? Can I identify the assumptions behind this design? Can I investigate an unexpected result? Can I recognize when the generated solution is solving the wrong problem? Can I explain the decision to another engineer without repeating the model’s explanation?
I do not think we need to preserve the ability to manually reproduce every line a machine can generate. Engineering has always advanced through abstraction. We rely on compilers, databases, frameworks, and libraries because they let us work at a more useful level.
The responsibility is to understand enough to make sound decisions, verify outcomes, and respond when something goes wrong.
An occasional period without AI can help us see whether that understanding is growing alongside our output.
It can also make room for something broader than technical competence.
Keep human things human
Some of the value of work comes from the experience of doing it with other people.
Asking a colleague a question may take longer than asking a model. But the conversation can reveal how they think, what they have tried, and what they are worried about. It can create the opening for a relationship that would never emerge from a faster answer.
A team brainstorming together may produce fewer ideas than an AI can generate in a minute. Yet the discussion can build shared understanding. People hear one another’s priorities. Someone tells a story that changes the direction. Someone disagrees, and the disagreement reveals the actual problem.
A manager can use AI to improve the wording of feedback. The manager still needs to notice the person, understand the situation, form a judgment, and be present for the conversation.
Those activities deserve our attention.
Efficiency is one value among several. Trust, empathy, belonging, curiosity, and shared experience also matter. Their contribution may be harder to measure, but we feel their absence.
Some conversations are valuable because two people took the time to have them.
Creativity
There is also a particular satisfaction in discovering what we can create ourselves.
When a model can offer twenty directions instantly, it becomes tempting to begin every creative process by choosing from its menu. That can be useful. It can also mean our own first thought never gets the chance to develop.
What would we have drawn on the whiteboard before seeing the generated options? What connection might we have made from an experience outside work? What unusual idea might have survived if we had given it another ten minutes?
I want room for that kind of creativity.
Human creativity does not need to win a competition against AI to be worth practicing. Making something has value for the person doing it. There is pleasure in finding a phrase, solving a problem, composing a melody, or building an object that did not exist before.
There is also something valuable about sharing the result with another person and saying, “Look what I made.”
AI can expand the ways we create. We can also choose moments when we want to experience more of the process directly.
Judgment
The same applies to judgment.
AI can organize evidence, suggest alternatives, and help us examine a decision. Someone still has to take responsibility for what matters in the specific situation.
Which tradeoff are we willing to make? Whose needs have we overlooked? What risk is acceptable? What promise are we making to a customer or a colleague? What kind of team are our choices creating?
These questions involve values, relationships, experience, and consequences.
Human judgment is imperfect. We should challenge it and improve it. AI can help us do that. But reviewing a recommendation should involve more than accepting a plausible answer.
An AI-free period offers a chance to state our reasoning first: what we believe, why we believe it, what remains uncertain, and what would change our minds.
That gives us something meaningful to compare with the model’s advice when we return to it.
The pace of work
All of these benefits matter. But the most immediate reason I am interested in AI-Free Fridays is the pace of work itself.
There is always another thing we could generate.
A prototype leads to another variation. A launch produces data to analyze. The analysis reveals opportunities. The opportunities become experiments. The experiments create more work.
That can be exciting. It can also make completion feel temporary to the point of disappearing.
Very quickly, “we can do more” becomes “we should do more.”
Yesterday’s extraordinary output becomes today’s baseline. The team adapts. Expectations rise again. Everyone is moving, but fewer people feel that they have arrived anywhere.
We can accomplish more while having less time to feel accomplished.
That is a strange outcome for technology intended to help us.
In the past, some pauses happened naturally because the next step took time. Those pauses were not always pleasant or useful. I do not miss waiting for slow tools.
But as more waiting disappears, we need to decide which pauses to create intentionally.
Making room to celebrate
A completed project deserves more than being moved to another column before the next task starts.
People deserve time to tell the story of what they did: the problem that looked impossible, the failed approach that taught them something, the careful decision that prevented trouble, the colleague who helped when they were stuck.
That is how an accomplishment becomes part of a team’s shared memory.
Celebration does not have to mean a large event, an award, or a polished presentation. Sometimes it means looking at what shipped and giving it our full attention. Sometimes it means asking someone to explain the part they are proud of. Sometimes it means saying, specifically, “I saw what you did, and it mattered.”
A generic congratulations passes quickly. Recognition becomes more meaningful when it shows that someone understood the work.
The judgment behind a simple design deserves recognition. So does the person who made a difficult migration safe, mentored a colleague, challenged a weak assumption, or decided that a feature should not be built.
As generated output becomes easier to produce, we need to notice contributions that cannot be captured by counting artifacts.
An AI-free period could give those contributions room to become visible.
We could spend time with something the team already built. Ask how it works. Hear what was difficult. Recognize who helped. Let people enjoy being good at what they do.
That was hard. We did that. And it was worth doing.
There should be room for that sentence before “What’s next?”
What leadership has to support
For this idea to work, leadership has to support the time it requires.
| If | Then |
|---|---|
| A team sets aside Friday afternoon for learning and reflection but keeps the same delivery expectations | The work may simply move into Thursday night or the weekend. The calendar changes while the pressure remains. |
| Recognition becomes another required presentation | It can add to the burden it was supposed to relieve. |
| People believe they will be judged for producing less during protected time | They will keep producing. |
An AI-Free Friday needs an actual decision about priorities. Managers must be willing to protect some of the capacity that better tools create.
That means making learning, conversation, and reflection legitimate uses of time. It means accepting an unfinished experiment. It means letting a discussion be useful without demanding a document afterward. It means allowing people to enjoy an achievement without immediately turning it into a new target.
Turning off a tool cannot repair an unhealthy pace by itself. The expectations around the tool have to change too.
A modest experiment
I would begin with a modest experiment.
- Investigate something they want to understand
- Practice a skill
- Work through a design with a colleague
- Create something from scratch
- Revisit an accomplishment that passed too quickly
- Ordinary toolsDocumentation, search, debuggers, calculators, and conversations.
- AccessibilityTools that enable someone to participate remain available.
- OperationsIf production is down or a customer needs urgent help, use the tools that help resolve the problem.
- What did people notice?
- Where did the restriction add no value?
Would miss the pointA rigid ban, a surveillance exercise, or a competition to prove who needs AI least.
Set aside a Friday afternoon, or even two hours. Let people choose a worthwhile activity: investigate something they want to understand, practice a skill, work through a design with a colleague, create something from scratch, or revisit an accomplishment that passed too quickly.
Keep ordinary tools available. Documentation, search, debuggers, calculators, and conversations can all support the work. The focus is on temporarily setting aside generative AI where doing so creates a useful experience.
Keep accessibility and individual needs in mind. Tools that enable someone to participate should remain available. People should not have to disclose a personal need to justify an exception.
Operational responsibilities also continue. If production is down or a customer needs urgent help, use the tools that help resolve the problem.
A little flexibility makes the experiment more useful.
The goal is to create protected space people can actually benefit from. A rigid ban, a surveillance exercise, or a competition to prove who needs AI least would miss the point.
After a few weeks, ask what people noticed.
Did they understand something more deeply? Have a conversation they would otherwise have skipped? Find a task they enjoyed doing themselves? Discover a skill they wanted to develop? Finally appreciate an accomplishment that had disappeared into the backlog?
Also ask where the restriction added no value.
Some tasks may become tedious without providing any learning or satisfaction. Others may turn out to be far more enjoyable when done directly. Those differences are useful information.
The experiment could make our use of AI more deliberate.
We might return to a tool with a better understanding of why we want it. We might choose to form an initial view before asking for critique, attempt a problem before requesting help, or talk to a colleague before generating another document.
We might also discover that the most valuable change had little to do with the tool.
Perhaps people simply needed permission to slow down together.
The ability to choose
There is an obvious irony in developing an argument like this with AI assistance. I am comfortable with it. Using a tool extensively can be part of understanding where we want boundaries around it.
We do not need a claim of purity. We need the ability to choose.
AI-Free Fridays might become a whole day, a short weekly practice, or an occasional team experiment. Different people and professions will find different formats useful.
The principle is that greater capability should create room for more than greater volume.
I want AI to help us build ambitious things. I also want it to create room for someone to learn a new skill, explore an uncertain idea, have a thoughtful conversation, and feel proud of a difficult piece of work.
We should be able to spend some of the time we gain on being human.
The work will keep coming. The next opportunity will remain visible. There will always be another prompt we could write.
Sometimes we should leave it unwritten for a while.
Think through the problem. Talk to the person. Try something we cannot yet do. Make something ourselves. Appreciate what the team accomplished.
Doing these things changes us.
That is part of their value.
And when we finish something meaningful, perhaps we can give ourselves enough space to enjoy it before asking what comes next.