Companion piece to Park Howell’s version at Business of Story. Same theme, two vantage points: Park writes from the brand side, I write from the operator’s seat.
Here is the uncomfortable truth at the center of the AI adoption crisis: the professionals most paralyzed by AI aren’t being robbed by it. They’re committing the theft themselves.
Every disruption of the past fifty years — internet, mobile, cloud — threatened what you do. Your distribution. Your infrastructure. Your workflow. You retooled, and your professional identity survived intact.
AI is different. It doesn’t threaten what you do. It threatens who you are. The architect whose career is built on knowing how enterprise systems talk to each other. The CFO whose seat rests on judgment compounded over twenty years. The operations leader who is the only person who knows how the company actually runs. AI doesn’t threaten their processes. It threatens to commoditize what they spent decades becoming.
So they respond the way people protecting an identity respond. Some freeze — they don’t touch the tools at all, because every experiment is a chance to be seen failing at the thing that defines them. Some hedge — they use AI quietly but never advocate for it. And some convince themselves that if they just keep their heads down, this will pass like every other hype cycle.
It won’t. And the research says the fear driving all three responses has nothing to do with the technology.
The Fear Has a Name, and It Isn’t “Robots”
A 2025 peer-reviewed study in ScienceDirect found workers describing “the conviction that an opaque algorithm will produce an error, and that the human operator will shoulder the blame.”
Read that again. Not fear that AI will replace them. Fear of being seen making a mistake with it. A field experiment found workers deliberately ignoring AI recommendations when managers could observe them — knowingly choosing worse outcomes to protect the appearance of independent expertise.
That is the actual adoption crisis. Not capability. Image.
And it inverts the org chart in a way most leaders miss: the most qualified people in your organization have the most identity invested in not being seen as wrong. Your best people are often the ones slowing you down most — whether that looks like quiet avoidance, a governance committee on its forty-third meeting with nothing in production, or a months-long internal build that exists mainly so nobody has to admit someone else could do it faster.
There is only one exit from this trap, and it isn’t waiting, hedging, or delegating the problem to a committee. It’s learning the instrument. Leaning in. Making the mistakes on purpose, in the open, early — while they’re still cheap.
You Already Know How to Do This
Here’s the reframe that unlocks it, and it’s sitting in plain sight: you already manage intelligence that isn’t yours.
When a leader delegates work to a direct report, the work changes hands but the accountability doesn’t.
You review the output. You catch what’s wrong. You send it back with direction. When it ships, your name is on it — and nobody thinks the existence of your direct report makes you replaceable. Directing capable people is the job. It’s what the judgment you spent twenty years building is for.
AI is exactly the same paradigm. It’s a capable, tireless, occasionally wrong junior team member. You delegate to it, you review what comes back, you correct it, and you remain accountable for what ships. The skill it demands — knowing what good looks like, catching the subtle miss, giving direction that improves the next draft — is precisely the skill senior people already have. That’s why the fear is backwards: AI doesn’t commoditize twenty years of judgment. It’s the first tool that lets twenty years of judgment operate at full throughput.
Nobody expects a new manager to delegate perfectly on day one. You learn it by doing it — by handing off work, getting back something wrong, and getting better at the handoff. Learning to direct AI is the same muscle, learned the same way: through reps, most of which will be imperfect. The leaders refusing to take those reps aren’t protecting their expertise. They’re refusing the promotion.
Green Irony’s AI-First CV
I’ll show you ours, because the credentials on it aren’t the wins. They’re the mistakes.
In 2024, we bet the company on AI-native delivery — we rebuilt every internal process around digital labor before selling that capability to a single customer. Most of those rebuilds didn’t work the first time. Many didn’t work the second. Failed prompts. Workflows that broke in ways that exposed processes which never needed to exist. Outputs that were confidently, instructively wrong — and got reviewed, corrected, and re-run, the same way a leader handles a junior hire’s first drafts.
Two years in, here’s what that CV reads like: we run on Claude and Notion. A single founder-operator directs what used to be separate executive functions — strategy, sales, marketing, delivery, parts of finance and legal — through an internal operating stack built one mistake at a time. Operating expense is down roughly 75%.
That’s not a productivity gain. That’s a different shape of company.
And the most important line on the CV isn’t a number. It’s a lesson: mistakes only compound if they have somewhere to land.
Early on, our failed experiments evaporated. A prompt that didn’t work taught the person who ran it and nobody else. The unlock was building what we now call an AI context layer — a governed, structured layer of institutional knowledge the AI actually operates from. Every decision, every correction, every “we tried this and here’s why it broke” gets written down where the AI can read it. Claude does the work; the context layer holds the judgment.
That changed the economics of failure completely. A mistake captured in the context layer makes the whole system permanently smarter — the review you gave yesterday’s output is baked into tomorrow’s. A mistake that lands nowhere is just a cost. Most organizations experimenting with AI are paying full price for their mistakes and capturing none of the learning. That’s why generic AI feels flat: the intelligence isn’t missing. The context is.
Culture Is the Technology
A study of 2,257 employees confirmed what we lived: psychological safety is the single strongest predictor of whether employees adopt AI at all. The technology is secondary. The culture is primary.
An AI-first culture isn’t a policy document or a tool budget. It’s the most senior person in the room sharing a failed prompt in a team meeting. It’s “I tried this, it didn’t work, here’s what I learned” — modeled from the top, about their own work. When leadership treats AI mistakes as curriculum instead of embarrassment, the whole org starts taking reps. When leadership performs finished expertise, the whole org hides — and the learning your company needs walks out the door to a competitor who allowed it. McKinsey pegs the prize at $4.4 trillion annually; the organizations capturing their share have one shared trait — they normalized experimentation before their competitors did.
Where to Start
Pick one process you personally run. Not your team’s. Yours. Delegate it to AI this week the way you’d delegate to a promising new hire: hand it off, review what comes back, correct it, run it again. Do it in the open. Write down what failed and why — somewhere your AI and your team can see it.
Five hours a week of executive attention on your own desk, for sixty days, will teach you more about what AI does for your organization than a year of committee output. Not because the experiments will all work — because the ones that don’t are how you learn what to build next, and every one of them starts compounding the moment it lands in your context layer instead of evaporating.
The leaders who win this era won’t be the ones who protected their identity from every visible mistake. They’ll be the ones who traded the performance of finished expertise for the identity that actually compounds: the person who learns faster than anyone else in the room — and built the system that remembers it.
The biggest mistake you’re making with AI right now is not making enough of them.
The second biggest is believing those mistakes say something about who you are. They don’t. They say something about who you’re becoming.
If you’re a mid-market or SMB leader figuring out what your operating model looks like on the other side of an AI-native rebuild — or you need AI-accelerated MuleSoft or Salesforce implementations on fixed-bid timelines — Green Irony builds working systems in weeks, not months, and runs executive AI advisory engagements that start on your desk, not your org chart. Start the conversation at greenirony.com. For the brand-side view of this same argument, read Park Howell’s companion piece at Business of Story.
Sources
- Frenzel, A., et al. (2025). “Guiding Employees to Embrace AI at Work.” ScienceDirect. sciencedirect.com
- Almog, D. (2025). “Barriers to AI Adoption: Image Concerns at Work.” arXiv. arxiv.org/abs/2511.18582
- “Safety First: Psychological Safety as the Key to AI Transformation.” arXiv. arxiv.org/abs/2602.23279
- McKinsey Global Institute. Superagency in the Workplace. mckinsey.com