The ancient symbol of the ouroboros shows a snake swallowing its own tail. It stood for endless renewal. In your organisation, it may soon stand for something less comforting.
When calculators arrived, we stopped doing long division by hand, and nobody missed it. Someone still understood the maths. AI is different. It doesn't take over one narrow skill. It takes over drafting, analysis and a growing share of the critical reasoning: the struggle that used to turn capable people into people with judgement.
Performance isn't capability
A 2025 field experiment published in PNAS gave nearly 1,000 high-school maths students access to GPT-4 during practice. Grades jumped: 48% for a standard ChatGPT-style tool, and 127% for a version designed to protect learning.
Then the researchers took the AI away. The standard-tool group scored 17% worse than students who never had access. The tutor group performed similarly to students who never had access.
Yes, I recognise these were teenagers, not executives (although this observation hasn’t affected Growth Mindset!) But the mechanism is familiar to any workplace: without guardrails, people lean on the tool as a crutch and then do worse on their own.
Your dashboards measure the first number, but your future depends on the second.
Where the snake bites
In a study of more than 5,000 customer-support agents, AI assistance lifted productivity by about 15% on average. Less experienced workers improved in both speed and quality. The most experienced saw small gains in speed and small declines in quality.
To be fair, the researchers also found signs that the tool helped newer agents learn. That tool suggested responses and left the decisions with the agent. Hold that thought…
Now recall my ‘Vanilla’ blog and what the original leadership research said develops leaders: “tough jobs, and the boss”.
In 70:20:10 terms, AI is starting to absorb much of the 70, stand in for parts of the 20, and generate the 10 on demand.
The junior staff gaining most today are tomorrow's managers, so if AI takes on the struggle, where does their judgement and critical thinking come from?
The edge you can't see
A Harvard-led experiment with 758 BCG consultants found that GPT-4 made them faster and better on tasks inside AI's capability. On a task just outside it, consultants using AI were 19% less likely to reach the right answer.
The catch is that the boundary is invisible unless you already have judgement, judgement built through the very struggle that AI removes. That's the snake eating its tail.
“So do we ban it?”
No! Both studies point the same way. AI that coaches, or suggests and leaves the decision with the person, builds capability. AI that hands out answers can erode it. The difference came from design, not the technology itself. That makes it a decision about where your budget goes, not a decision for IT.
Four questions before your next AI rollout
Does this tool give answers, or make people think?
What happens to performance when it's switched off?
Where will your next managers get the hard experiences AI now absorbs?
Who can tell when the AI is past its limits, and how are you developing more of those people?
If the room goes quiet, the snake has already started eating.
Where Plexity Partners fits
Plexity works where business acumen and people performance meet. I sit as an independent partner on the buyer's side.
Use AI to sharpen the brief, but call me to decide what the brief should be: where capability investment builds judgement, and where it quietly erodes it.
Coming next: The AI capability investments that build judgement, and the ones that quietly replace it.
References
Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakcı, Ö., & Mariman, R. (2025). Generative AI without guardrails can harm learning: Evidence from high school mathematics. Proceedings of the National Academy of Sciences, 122(26), e2422633122. https://doi.org/10.1073/pnas.2422633122
Brynjolfsson, E., Li, D., & Raymond, L. (2025). Generative AI at work. The Quarterly Journal of Economics, 140(2), 889–942. https://doi.org/10.1093/qje/qjae044
Dell'Acqua, F., McFowland, E., III, Mollick, E. R., Lifshitz, H., Kellogg, K. C., Rajendran, S., Krayer, L., Candelon, F., & Lakhani, K. R. (2026). Navigating the jagged technological frontier: Field experimental evidence of the effects of artificial intelligence on knowledge worker productivity and quality. Organization Science, 37(2), 403–423. https://doi.org/10.1287/orsc.2025.21838
McCall, M. W., Jr., Lombardo, M. M., & Morrison, A. M. (1988). The lessons of experience: How successful executives develop on the job. Lexington Books.