AI, fake productivity, and why “time‑tracking founders” are in trouble
And why I am happy with our approach
Recently I had a weird moment.
I was sitting in an internal Welltory workshop. One of our engineers was showing the team how to build plugins for Claude Cowork. It’s a powerful thing: you can automate a scary amount of your own work. People were watching and you could almost hear the collective gulp — oh, shit, this just got real.
You could see the gears turning:
“Wait. I can wrap my whole operational hell into a couple of pipelines and never touch it again?”
At some point someone joked:
“If we had time tracking here, we could now fake being busy forever and do nothing.”
My first reaction was a small internal panic.
The second was relief.
Thank God we never tried to control hours.
The next thought was less kind:
Oh wow. So many companies are already screwed and don’t know it yet.
For years people kept telling me:
you can’t do real innovation remotely
you have to watch people
you need to “look them in the eyes” in the office
remote only works for juniors who need supervision
Now, in 2026, it’s pretty obvious what actually matters for any company building an intellectual product — and what doesn’t. And everyone who bet their culture on controlling presence and screen time just lost that bet.
Let’s rewind a bit and see how we got here.
We were paying for theater long before AI showed up
There’s an ugly fact we have to admit: the “activity instead of results” problem is not new.
A Connext Global study found that about two thirds of U.S. employees openly admit they engage in productivity theater: they pick up “visible” tasks, stay late, and stay active in chat mainly to look busy. Only around 23% say their performance is measured against clear outcome metrics, not how hardworking they appear.
Sources:
So tens of thousands of companies have been paying not for work, but for rituals around work: speed of email replies, number of calls, how pretty the slide deck looks.
And then AI walked in.
A culture built on hours meets AI
Picture a company where the whole culture is built on time:
efficiency = how many hours you “sat at work”
processes are built around presence, not results
management is basically surveillance: who arrived when, who stayed how long, who is green in Slack
Now drop modern AI into this setup:
Claude / coding agents that can hold basic work conversations and operate tools for you
agents that move files, create tasks, draft reports and summaries
tools that fill in forms and documentation better than the most patient junior
(Examples of these:
Cursor & Claude agents overview
Claude agents breakdown
The whole “control by hours” logic collapses in one release.
Mouse movement? Old‑school mouse jigglers were already a thing years ago. Employers now publish long guides on how to detect artificial cursor movement by speed and acceleration patterns:
VanHack on spotting mouse jigglers
EmpMonitor on detecting jigglers
On top of that, you can glue an AI agent that moves the mouse like a human and “uses” apps all day.
Being active in chat? A bot can answer “got it, will check” and “let’s talk tomorrow” just as well as half your org.
A 10‑page report by 6 p.m.? What used to be a full day of work is now ten minutes of prompting. Field reports and vendor data show double‑digit productivity gains from AI on writing, research, and coding tasks, especially in knowledge work.
Overview example.
If your main metric is visible activity, AI is a machine for synthetic activity.
Cheaper, faster, and tireless.
Life inside a “visibility corporation”: the numbers
Let’s put some numbers on this.
Microsoft’s Work Trend Index shows employees spend up to 57% of their time in meetings, email, and chat — and only 43% on focused work that produces value.
See here and at a summary
Meeting time has more than doubled since 2020, but average meeting length barely shrank.
Nice digest here.
Focus research from Gloria Mark’s group finds that after a serious interruption, the brain needs about 23 minutes to fully get back into the task. One badly placed meeting in the middle of a deep work block can wipe out half an hour of real productivity.
Classic paper: “The Cost of Interrupted Work: More Speed and Stress”
On top of that, “bossware” is quietly spreading: screen‑capture tools, keyloggers, app trackers. Legal and HR folks are already writing whole reports about it — and they all come to the same conclusion: surveillance erodes trust, spikes stress, and increases turnover intentions.
Примеры:
On “bossware” in general
Academic work on monitoring, trust, and turnover
Against this backdrop, “let’s strengthen control” is not a plan.
It’s the corporate equivalent of treating a migraine with a hammer.
The unwinnable arms race
The “we’ll just watch them more closely” path sends your company into a pointless arms race:
on one side, monitoring vendors invent ever more sensitive ways to flag “abnormal” patterns
on the other, AI agents get better at mimicking human behavior and producing perfect “busywork” traces
Examples of how fast agent capabilities are evolving:
Security folks red‑teaming coding agents and showing how they can be scripted
Who wins a game of “who’s sneakier”?
The side that has AI and motivation to use it. That’s not your middle manager writing reprimands because “you were red in Slack for too long”.
Meanwhile:
faking activity gets cheaper
separating signal from noise gets harder
managers drown in dashboards about “engagement” and “time in app” and still have no idea what actually moves the business
Worklytics and others keep pointing out this trap: leadership obsesses over activity metrics instead of outcome metrics, and then wonders why nothing moves.
Example KPI piece:
Productivity benchmarks:
It’s not just pointless. It’s expensive. You’re burning money on:
control software
payroll for people playing cat‑and‑mouse with that software
and you’re not investing in the only things that actually raise productivity: focus time, sane AI‑supported workflows, and clear goals
Deep‑work stats for context:
Deep work & focus time trends
Knowledge worker productivity & focus time
The companies that look strangely calm
If you look at the few places that seem oddly calm about AI, they have one thing in common:
they shifted to managing by outcomes long before the current model hype.
Netflix famously has no vacation tracking and minimal formal rules. Their culture memo basically says: “No vacation policy. No approval process.” The flip side is brutal clarity on performance and “talent density”: you are trusted like an adult and measured like an adult.
- Netflix culture memo
- Book / context
- A good summary
GitLab runs one of the largest all‑remote organizations in the world with a fully public handbook. They don’t track hours; they track impact and documented outcomes. Their leadership talks openly about the trap of rewarding people for presence and responsiveness instead of real contribution.
- All‑remote culture
- Remote benefits
- Nice external case write‑up
The older ROWE (Results‑Only Work Environment) experiments at Best Buy showed the same pattern: when they stopped tracking time and switched to pure results, voluntary turnover dropped dramatically and productivity jumped.
Intro to ROWE:
Case discussion
Productivity analytics across thousands of knowledge workers shows a simple thing: the people who reliably get 3–3.5 hours of deep focus per day outperform those who live in “meeting‑chat‑meeting” mode.
Benchmarks here.
Deep work stats here.
Bossware doesn’t create deep focus. It destroys the little that’s left.
So no, the answer to AI is not “more screenshots”.
OK, but what do you actually do instead?
“Don’t track hours, manage outcomes” sounds nice on a panel.
It’s less nice at 10 a.m. on Monday when you have a real company to run.
Here’s what I’d do if I were trying to turn a time‑tracking culture around.
1. Translate your company into outcome‑speak
As long as your org speaks in tasks — “run the project”, “support the system”, “do marketing” — any talk about trust will stay abstract.
Outcomes look more like this:
not “handle support”, but keep NPS above X and first‑response time under Y
not “do marketing”, but bring in X qualified leads at ≤ Y CAC
not “write code”, but move retention, ARPU, or CSAT by X points this quarter with specific product changes
Worklytics and similar firms publish KPI lists for knowledge workers that are all about contribution and impact, not about time in front of a screen:
KPI examples
They’re a good starting point if you’re staring at a blank page.
2. Build AI into the job, not as a side hobby
If you treat “real work” as manual and AI as a cheat code, you’ve already lost.
Instead:
write into role descriptions: “can design workflows that use AI/agents, not just do tasks by hand”
ask in performance reviews: “Which parts of your work have you already automated? What’s next?”
measure people not by “time spent”, but by how much output improved versus the previous period
McKinsey, IDC and friends keep repeating this point: the winners won’t be “the ones who bought AI”, but the ones who rewired their work around humans + agents working as a system.
Some representative takes:
McKinsey post on agentic AI giving employees “superpowers” (via FB share)
a16z on AI workflows
3. Invest in outcome‑thinkers, even if they’re AI‑illiterate today
This is where many leaders make the wrong cuts.
If you have someone who:
genuinely cares about the business result
thinks in outcomes and takes ownership
but doesn’t yet know how to work with AI,
this is your best investment.
The simplest training pattern I’ve seen work:
Give them a task that can’t reasonably be done without AI — not a toy prompt, but a real piece of work that would be too big or too tedious by hand.
Say out loud:
“The first time you do this with AI, it’ll be slower than doing it manually. That’s okay. The point is to build the muscle for the next ten tasks, not to beat your current speed on day one.”Show where the help lives: internal AI champions, workshops, a short curated list of courses or docs.
OECD and the IMF both point out that companies actively investing in AI‑related skills are more likely to see productivity gains and less likely to lean purely on job cuts:
In SME surveys, significantly more firms say AI has raised skill requirements than lowered them.
If someone can already think in outcomes, they almost always figure AI out — especially if you stop punishing them for not being “faster than manual mode” on day one.
4. Be honest: you will part ways with some people
Here comes the hard part.
You will almost certainly have to let go of some folks:
those who refuse to think beyond checklists (“just tell me exactly what to do”)
those who see AI only as a threat and won’t touch it
those whose core skill is “being around and looking busy”
This isn’t about IQ or elite degrees. It’s about a shift in what the job is. From “execute instructions” to “hold the goal in your head and assemble a system — of people and machines — that gets there”.
OECD and IMF both point to a growing demand for hybrid roles where humans set direction and make calls, while routine execution gets eaten by software:
If your org is full of people who only want the execution part, AI will come for that.
Trying to preserve a workforce that was bred for process and presence control is, in 2026, the HR equivalent of fighting to keep your fax department in 2005.
5. Realize what the talent war is now about
This, to me, is the core of the whole thing.
The war for talent is shifting toward people who:
see the end business outcome, not just the next ticket
can break that outcome into steps, some for themselves, some for AI/agents
accept responsibility for the outcome, even if half the work was done by non‑humans
You can’t start competing for these people “later, once things settle down”. By then they’ll already be working somewhere that gave them trust and interesting problems.
Right now:
AI tools are basically electricity — everyone can plug into them
(good overview of how ubiquitous they already are)The real differentiator is not “who has ChatGPT”, but who has people who can orchestrate AI into working systems
(see e.g. a16z)The companies that are already stripping out time‑tracking and rebuilding around outcomes are going to look freakishly efficient in a couple of years next to those still buying fancier bossware and wondering why the margin doesn’t move.
A personal note
At Welltory we don’t track hours.
We have unlimited paid time off, sick days, and weekends.
We measure a person’s value by their contribution and what they do for the team — not by how long they sit in Zoom.
So when someone in our workshop jokes:
“With time tracking, we could now stop working and just let AI fake it,”
we can laugh. There’s nothing to “cheat” in our time‑tracking system. There isn’t one.
For companies built on control, it’s not a joke.
It’s a business risk and a cultural dead end.
Because at some point you have to admit:
you’ve been optimizing the wrong thing for years
AI just poured rocket fuel on that mistake
and now you either rebuild around outcomes and trust, or watch your own employees train agents to fake work better than they ever could themselves
If you’re a founder or a leader, your choice is pretty stark:
double down on control — and lose to your own AI in your own game
or double down on outcomes and trust — and start competing for the people who can actually keep your company alive in an AI world
I know which side I’m on.