💯 Digest: OpenAI admits AI can't replace you (yet)

Window into the Future 🔮:

Why you should care (not panic) about OpenAI’s new report

OpenAI just put out a 20+ page report called GDPval that measures AI's real economic impact (aka measuring whether AI can do your job and hence the GDP in the name). They hired professionals with 14+ years of experience from 44 different jobs across 9 sectors that collectively earn $3T annually.

The TLDR is AI can supercharge your work today, but only if you stay in the driver’s seat.

So let me back up and tell you what this report says about your job security (spoiler: OpenAI literally shows that their competitor is better at the work you actually do)

What They Actually Tested

They gave both humans and AI models the same real-world tasks: create presentations, analyze data, write reports, and design materials. The kind of stuff that fills your Monday mornings and Thursday afternoons.

Here are the bits that stood out to me:

• AI outputs were as good as or better than human professionals in just over 50% of cases.

• OpenAI's own study shows Claude beating GPT models at actual workplace tasks. GPT-5 was strongest on accuracy (correct math, careful instructions), while Claude Opus 4.1 led on style and presentation (slides, layouts, design).

• Using AI as a first draft tool (with a human reviewing or fixing) cut costs by up to 63% and time by up to 39%.

• When letting the AI do the work without review, tasks were completed 90x faster and hundreds of times cheaper, but with lower quality.

• Left unchecked, the risks are real: about 29% of GPT-5’s misses were rated “bad or catastrophic,” and 3% were flagged as dangerous (like wrong medical or financial advice). Most slip-ups, though, were boring but common formatting errors or half-finished deliverables.

• Simple tweaks like prompting GPT-5 to “double-check formatting” or “review deliverables” actually improved performance and cut down errors.

• The average task took a pro 7–9 hours to complete, sometimes stretching into weeks, with an average “work value” of around $398 per task.

My takeaway is that AI is already closing in on professional work in half the tasks tested, but it’s not reliable enough to leave unsupervised. If you use it as a co-worker who drafts, speeds things up, and saves you money (but still needs oversight), you can get real value today.

So let me say that again: It's not AI vs. humans. It's humans WITH AI absolutely crushing it.

Three things to remember as you're sipping your coffee tomorrow ☕️

• That monster voice in your head telling you that you'll be replaced by AI is looking at the wrong data. The real win is in getting rid of the boring and repetitive tasks (you shouldn’t have to start your presentation from scratch anymore).

• The professionals getting 10x results are people who've figured out the dance with AI: when to lead, when to follow, and when to take the wheel completely.

• Take a task you usually spend hours on and make AI do the messy first draft. Then spend your time reviewing, editing, and polishing. That’s the “human + AI” setup that the report found delivers the biggest payoff.

And that's exactly what we're teaching in both our free challenge and our bootcamps. How to think, build, and multiply your impact with AI.

Read. Watch. Listen. To... 📖

We’re not just learning new tools. We’re learning how to work differently. These pieces helped me think deeper (and faster):

• 📕 Learn how to use ChatGPT for any role

• 📺 OpenAI introduces Pulse

• 🐦 How more companies are defaulting to AI

Before You Go…

What are your thoughts on this report? 🤔 Have you checked it out, and if so, did you find any interesting insights that I might have missed?

Cheers!