Why the most urgent work of the AI era isn’t technical. It’s learning, finally, to see ourselves.
Here is the short version, before the argument: what an organization calls its shared operational truth is a constructed reality — a version of the world everyone inside has quietly agreed not to question. Agentic AI does not correct that construction. Built the way most organizations are building it, the AI learns the construction, acts on it, and multiplies it.
What this piece argues
- 01Every organization already runs on a constructed reality. Not its culture — the layer beneath it, which decides which numbers reach the dashboard at all.
- 02Almost nobody trains a model. What companies build is the harness around it — and a harness assembled from their own history reimports their true character at full fidelity.
- 03That makes an AI deployment an amplifier rather than a mirror. The work left to leaders is to decide, deliberately, what it amplifies.
The essay works through a morning in a fourth-floor room in Amsterdam, the neuroscience of how perception actually functions, and the one example everyone recognises themselves in — hiring. It ends with three things to do before the deployment roadmap gets signed.
Point a tireless, capable system at a constructed reality that nobody has examined and you have not built a mirror. You have built an amplifier.
Read the full perspective
Which reality is your organization about to encode?
We help leadership teams surface the operating truths their systems are about to learn from — before the roadmap is signed.
Start the conversation