1. 90% of Executives Say AI Hasn't Boosted Productivity

On August 22, Fortune published research from University of Pittsburgh professor Mark Ma that should make every leader pause.

Citing an Atlanta Federal Reserve study, Ma notes that about 90% of executives believe AI has not yet boosted productivity at their companies. Then he went looking for why.

His team analyzed millions of Glassdoor reviews, thousands of financial reports, and hundreds of AI investment and layoff announcements from US public companies over five years. The pattern was clear: as AI investment announcements go up, so do announcements of AI-driven job cuts. And those cuts appear to be the thing quietly destroying the returns.

Here's the mechanism.

Employees are told to adopt AI tools while watching colleagues lose jobs to those same tools. AI-related comments in employee reviews came back much more negative than the overall tone of reviews, with job security fears the loudest complaint by far. And employee sentiment toward AI turned out to be one of the strongest predictors of firm productivity.

Management optimism, meanwhile, predicted nothing. Ma's team analyzed the tone of roughly 10,000 earnings call transcripts and found executives consistently sunny, with no significant relationship to actual productivity outcomes.

Investors noticed. The average stock reaction to AI-cited layoff announcements was close to zero.

The takeaway

How your team feels about AI matters more to your ROI than how you feel about it. If you want the productivity gains, share them. Invest in skills and expand opportunity instead of using the AI budget as a justification for headcount reduction. The companies treating this as a capability problem are the ones getting paid.

2. ChatGPT Work Can Now Trigger Itself

On August 25, OpenAI shipped two ChatGPT Work updates worth ten minutes of your afternoon.

Webhook-triggered tasks. Scheduled tasks no longer have to run on a clock. They can now fire when something actually changes: a new Gmail message arrives, a Slack channel gets a message, or a GitHub pull request updates. So instead of checking Slack for client feedback, you can have ChatGPT draft next steps the moment that feedback lands. Available on Plus and Pro, and you need to add @ChatGPT to each monitored Slack channel. Anything requiring approval pauses until you review it.

Tasks on logged-in sites. ChatGPT Work's browser can now complete work on websites that require a sign-in. It surfaces the login screen, you enter your credentials (password managers are supported, and OpenAI says the model never sees or stores your username and password), and it keeps working while you step away. It asks for confirmation before consequential actions like payments or bookings.

You can also now share a scheduled task so a colleague can review it, customize it, and spin up their own independent copy with their own connected apps. Free users get up to three active scheduled tasks.

The catch

A TechCrunch feature on August 24 about OpenAI's agent push surfaced a stat that reframes all of this. An OpenAI-backed study found that in June, nearly 98% of OpenAI employees were using Codex (ChatGPT’s AI coding agent). Among organizational subscribers, it was 17%. Among individual subscribers, under 1%.

The tools are getting more capable much faster than teams are learning to use them. That gap is where your advantage sits right now, and it costs $20 a month to start closing it.

3. An Easy Way to Turn Data Into Profits

This week’s newsletter is sponsored by StoreClaw. More details below.

AI Insights That Turn Your Data Into Bigger Profits

Your store is full of hidden growth opportunities. StoreClaw analyzes your Shopify and Amazon data, identifies the highest-impact actions, and helps you grow revenue while improving margins. Start free today with 10,000 bonus tokens. No credit card required.

Try it FREE today

4. Uber's $966M Lesson on Automated Decisions

On August 21, the Dutch Data Protection Authority fined Uber €825 million, about $966 million, for deactivating driver accounts through an automated process without sufficient warning or human oversight. It is the second-largest penalty ever issued under GDPR.

Deputy chair Monique Verdier's framing is the line to remember: a computer should not be making decisions on its own that carry consequences that large.

To be fair, Uber disputes the finding. The company says most suspensions are brief, that no permanent deactivation happens without human review, and that drivers can appeal. It is appealing the decision. There's also a legitimate counterargument circulating that blaming "a computer" is a bit like blaming the time clock when a company fires a chronically late employee. Managers set the policy; the system just measures compliance.

But the regulatory direction is what matters for your planning. This is the third Dutch fine against Uber on related issues, and the complainants are now organizing a class action.

The takeaway

Find every place in your organization where an automated system is making, or effectively making, a consequential decision about a person. Hiring screens. Performance flags. Account suspensions. Contractor deactivations. Each one needs a documented human review step and a real appeals path, not a rubber stamp.

Do this audit before someone asks you for it. "The model decided" is not going to hold up.

5. One Prompt You Can Use at Work Today

Given what the research in the first section found, here’s a prompt that can help you at work:

For example,

P.S. I’m giving a free talk next week on Udemy about “Breaking the AI Silo: Building a Culture of Shared, Visible AI Practice.” Click here to watch the live webinar on September 1 (it’s free).