Everyone is talking about the race to AI. The race to automate. The race to improve productivity. The race to prove return on investment.
There is a temptation to believe that the competitive advantage is in being faster, but many of us know the fable that challenges that assumption. The tortoise and the hare. We remember it as “slow and steady wins the race.” I think there’s more to learn from that story, and two recent events show why.
The Hare’s Real Mistake
The hare didn’t lose because he was fast. Speed was never his problem. Once he’d built up a big enough lead, he stretched out by the side of the road for a nap, confident nothing could catch him. He treated the finish line as the only thing that mattered, so he stopped paying attention to the steps in between. He assumed his lead was enough, and he had no way to prove otherwise until he woke up and found the tortoise already there.
The tortoise won for a different reason than the one we usually tell. It wasn’t patience for its own sake. Every step he took was visible and accountable. Anyone watching could see exactly where he stood at any point in the race.
That’s the distinction I think matters right now. Some organizations are racing toward a finished result: how much work did AI complete, how fast did we get there, what’s the ROI. Others are building something incrementally, one visible, checkable step at a time. Those aren’t the same race, and this week gave us two clear examples of what happens when you run the first one and assume it’s the same as the second.
When the Finish Line Isn’t Enough
An Australian government department paid Deloitte roughly A$440,000 for a report reviewing the IT systems behind its welfare penalty program. It reached the department looking finished: polished, authoritative, ready to act on.
A Sydney University researcher found it full of fabricated citations, including a quote falsely attributed to a federal court judgment and references to academic papers that never existed. Deloitte revised the report, refunded part of the fee, and disclosed that a generative AI tool had been used in producing it (AP News).
Deloitte made the hare’s exact mistake. The hare got far enough ahead that he stopped worrying about anyone checking his pace, and coasted. Deloitte got a report far enough through a polished, credentialed process that nobody thought to check the pace of the work behind it either, until an outside researcher did what Deloitte’s own review should have done. Racing toward a finished deliverable and racing toward a trustworthy one are not the same effort, and Deloitte spent its energy on the one that only had to look right at the finish.
Losing Trust Without Losing Accuracy
Hank Green has spent nearly 20 years building one of YouTube’s most trusted platforms, as half of the Vlogbrothers but the creator behind SciShow and CrashCourse. He has more than 12 million followers and roughly 32 million subscribers across his channels, built almost entirely on the premise that his audience can trust what he tells them. His story is the subtler version of the same mistake, and I think it’s actually the more important one.
Green apologized to his millions of subscribers after fans noticed the phrase “I appreciate the pushback” in one of his videos, a stock phrase common in chatbot responses, and concluded he must have read from an AI-generated script. Green said that specific line was an ad-lib responding to a guest, not AI text, and on that narrow point, he was right.
But the apology didn’t stop there. Green also acknowledged using ChatGPT to prepare the video, describing it at first as research and generated notes, and conceded the finished piece had an “AI feel” to it. In a longer Reddit post, he explained he uses AI models to locate papers and source material, not to supply his conclusions or write the words he says on camera. He admitted he’d become too reliant on the tools regardless, largely because he’d overcommitted himself and was producing under pressure.
His audience didn’t fully accept the distinction, even though nobody proved his actual conclusions were wrong. Many viewers felt that AI-assisted research alone was enough to undermine the standard they expected from him, given ongoing concerns about hallucinated citations, weak source evaluation, and the mismatch between his reputation for careful, human-centered work and the workflow he’d just described. What broke down wasn’t accuracy. It was trust in the process behind the work.
Green admitted that his own process wasn’t clear to him anymore. He’d been moving so fast that he’d lost track of it himself.
Deloitte’s failure was a visible error somebody eventually caught. Green’s was in some ways more instructive. His actual output may have held up fine, yet the trust still collapsed, because the steps behind the work had stopped being visible to anyone, including him. He’d been running like the hare without realizing it. All finish line, no accountable steps along the way.
The Incremental Advantage
Put the fable and both stories together, and the real lesson comes into focus. The race to ROI treats the finish line as the whole point. Incremental improvement treats every step as something that has to hold up on its own, whether or not anyone happens to be watching.
You don’t have to be fast to win. You have to be deliberate. The tortoise just kept putting one foot in front of the other, the whole way, and that steady, visible pace is what made him someone you could count on before the race ever ended.
When people ask me how I’m using AI, I don’t have a lot of impressive automation stories to offer, even though I probably use it more than the average person does. I have information flowing to me automatically at the right moments, and I’ve built AI rules that keep me consistent and help me make decisions that stay aligned with my goals. What I’m not doing is rushing to build agents or looking for what I can hand off entirely. I focus on what I actually need and how AI can enhance it, and I stay involved in every step until I know I can trust what comes out the other end. It’s helped me fill in the gaps where I have known weaknesses, moving forward at a steady pace instead of a fast one.
That approach hasn’t made me faster. It’s made me incrementally better, and I think that’s worth considering at the organizational level too. Not how much work did AI complete, but how much better did our process actually get while we used it. Could someone else follow the steps we took and see real improvement, not just more output?
Three Checks Before Your Next Deliverable
“Legible process” is easier to act on than it sounds. Before your next AI-assisted deliverable goes out the door, three questions will tell you which race you’re actually running.
Can you trace who reviewed it. Not whether it looks reviewed. Whether a specific person can say what they checked and when.
Is AI’s role disclosed anywhere it materially shaped the outcome. Not buried in a policy document nobody reads. Visible to whoever is trusting the work.
Would the process survive someone asking you to walk them through how it got made. If the honest answer is “I’d have to reconstruct that,” the process was never actually legible. It just looked finished.
None of these slow work down in any meaningful way. They just make sure the steps that got you to the finish line could survive being looked at, which is exactly what neither Deloitte’s report nor Green’s research process could do when someone finally asked.
Those two questions build very different organizations. One optimizes for a finished result. The other builds something people can actually trust, one verifiable step at a time. And that’s the race that actually compounds. The organizations that win it won’t simply be the ones that automate the fastest. They’ll be the ones whose process stays legible at every step, not just polished at the finish line.
A note on how this was made: I used Perplexity for research and fact-checking, Claude to help organize my thoughts and work through structure, and Notebook to create the images. Each piece still takes me 4 to 6 hours to produce. The ideas, arguments, and words are mine.
Oh, and one more thing…
Navigate Change. Amplify Potential. features real leaders and real stories about navigating change and unlocking the potential of people. Through candid conversations with executives, entrepreneurs, and changemakers, we explore leadership, AI, innovation, workplace culture, and the human side of transformation. Episodes are broken into conversations short enough to watch on a break, but long enough to be worth your time. Give it a listen.