Note
The best AI agent may be the one you trust enough to let in
The best AI agent may not be the smartest one. It may be the one you trust enough to let in.
We’ve spent a lot of time comparing AI on capability. Agents make me wonder whether the more interesting constraint will be access.
A genuinely useful agent needs context. It might need your email, calendar, files, preferences or purchasing history. Inside an organisation that becomes identity, data, system permissions and potentially the authority to spend money or make changes.
An agent can be extraordinarily capable and still not be particularly useful if we aren’t prepared to give it the access it needs.
Which makes trust part of the architecture, not just a question for the privacy policy.
Capability requires access. Access requires trust. And trust determines how much of the capability we can actually use.
Perhaps the scarce resource in the agentic era won’t be intelligence at all. It will be trusted access.
Editorial decision (24 September 2026): publish on Threads, X and Bluesky; hold for LinkedIn pending a fuller organisational treatment. This is a draft, not a claim of publication.
Originally published on Threads.