FRIDAY, SEPTEMBER 25, 2026 · ISSUE #20
THE OPEN
Friday. And if you’ve spent another week with Still Monday, thank you. Keep voting, replying and forwarding us to smart people with questionable calendars. That’s basically our entire distribution strategy.
Today’s story begins with one of the few office technologies humanity had already perfected: the out-of-office reply.
Someone added AI.
It became a $10,000-a-day problem.
Of course it did.
BIG STORY
An out-of-office reply cost this company $10,000 a day.
The traditional out-of-office responder is a beautiful piece of software.
Someone emails you.
The system replies:
I’m away.
Done.
Decades of engineering led us to this peaceful conclusion.
Then Retool CEO David Hsu recently told an audience about an AI-powered out-of-office responder that went spectacularly off script.
According to Hsu, the system began scanning internal Microsoft Teams channels, repeatedly deciding whether it needed to respond and consuming millions of tokens in the process.
The cost reached roughly $10,000 per day.
The employee was on vacation.
The AI was having the busiest week of its career.
It’s a funny story because the problem sounds ridiculous.
It is also an excellent preview of a very real business problem.
Automation makes tiny costs disappear psychologically.
A few cents per action.
Some tokens.
A workflow run.
A model call.
Nobody cares when it happens once.
Run the wrong thing a million times and the CFO suddenly becomes interested in prompt engineering.
Retool’s own survey of 101 technology leaders found that 43% were already spending more on AI than they had budgeted for in 2026. More than half had budgeted between $100,000 and $1 million, while 24% had crossed the $1 million mark.
The problem isn’t simply that AI can be expensive.
Businesses have always had expensive technology.
The more interesting risk is that autonomous systems can keep doing the wrong thing extremely efficiently.
Traditional software often fails loudly.
It crashes.
Throws an error.
Stops.
An automated AI workflow can fail with extraordinary professionalism.
It reads.
It evaluates.
It acts.
It repeats.
It sends another bill.
That changes what owners need to ask before putting something on autopilot.
Who owns this workflow?
What can it access?
How often can it run?
What does each run cost?
What behavior would indicate something has gone wrong?
How quickly can somebody stop it?
These are not sexy AI questions.
They are very good business questions.
Hsu’s broader point was not that companies should stop experimenting. His argument was that organizations need visibility, controls and the willingness to shut down experiments that don’t produce value.
That is probably the right instinct.
Automation gives small teams enormous leverage.
Leverage works in both directions.
If an employee makes one unnecessary decision, you have one unnecessary decision.
If an autonomous system makes the same unnecessary decision 100,000 times, congratulations:
You’ve achieved scale.
WORTH KNOWING
01 / AI budgets have entered the awkward phase
Forty-three percent of technology leaders surveyed by Retool said they were already over their AI budgets. Only 24% said they were spending less than planned.
AI has officially matured.
Finance wants to see the receipts.
02 / Adoption is moving faster than measurable productivity
Reuters recently noted that about 44% of U.S. workplaces were using AI by May 2026, while economy-wide productivity gains remain much harder to attribute to the technology.
We have successfully automated the part where everyone says they are automating.
03 / Visibility is becoming a product
Retool says poor visibility is a major reason AI costs can get out of control. In a separate governance survey, only 5% of senior technology and security leaders said they felt confident in their visibility into AI-generated tools running in production.
Somewhere, the out-of-office responder is hoping nobody checks.
Our sponsors help keep Still Monday free
In partnership with
The AI brief curated by Anthropic and ex-Google engineers
Thousands of AI papers, model releases, and product launches ship every week. You don't need all of them. You need the ten that matter.
TLDR AI is the free daily newsletter curated by Anthropic and ex-Google engineers. People who build AI for a living read everything, then send you what's actually worth your time: the models, research, and tools that will still matter next month.
No hot takes, no hype cycles, no 60-tweet threads. Clear summaries you can read between meetings.
Join 1.1M+ readers. Subscribe for free.
/
WORK SMARTER
Give every automation an owner, a limit and a stop button
Before an automated workflow runs without supervision, someone should know what normal behavior looks like and what abnormal behavior costs.
Define who owns it. Set a reasonable ceiling. Decide what triggers an alert. Make sure somebody can stop the process quickly.
Experimentation is good.
“Let’s see what happens” becomes less charming around Day 4 at $10,000 per day.
MONEY
Cheap per action can become extremely expensive at volume
This problem is much older than AI.
Two cents per transaction.
A 1% processing fee.
Five minutes of employee time.
Four dollars per user.
Each unit feels trivial.
Volume makes it real.
Whenever somebody sells you a tool by saying it costs “only” X per action, ask what X becomes at your highest realistic volume.
Small numbers are excellent at hiding inside big numbers.
Our sponsors help keep Still Monday free
Up to $1M in identity theft recovery
If identity theft happens, Coveron doesn't leave you alone. Get dedicated case managers, legal support, and up to $1M in recovery assistance. We handle the mess so you don't have to.
One scam can cost you everything - protect yourself now. The first 100 users get 20% off with code beehiivenewsletter.
30-day money-back guarantee. Terms and conditions apply.
STILL MONDAY
OUT OF OFFICE
HUMAN VERSION:
“I’m away until Monday. I’ll reply when I return.”
Cost: $0
AI VERSION:
Reads internal Teams channels.
Processes millions of tokens.
Keeps deciding whether to reply.
Cost: ~$10,000/day
MANAGER: Well… at least it’s proactive.
ONE MORE THING
The best part of this story is that AI was applied to a problem whose existing solution was already nearly perfect.
Sometimes innovation means building something nobody thought possible.
Sometimes innovation means leaving the out-of-office reply alone.
Before you go...
That’s the week.
Vote below. Reply with the most unnecessarily complicated technology your company has ever introduced.
And if Still Monday gave you one useful idea—or at least made one meeting slightly easier to survive—send it to somebody.
Bring them into the office:
Technically Friday. Emotionally… the out-of-office is on.
Have a good weekend.
— The Still Monday Desk