A score of how much an agent acts can’t reach whom it was formed from. On the advisors’ Autonomy Score, and beginning at the person, not the account.

Conferred vs. Authored Meaning in the Age of AI On what AI dissolves rather than destroys, the question a pandemic asked and most people handed back unopened, and why private decision-makers reached this future first.

An agent can act on a customer's behalf and still not be on their side. On NAB’s customer-authorised agents, and which side of the table yours sits on.

This week an AI didn't just act on the account — it built the agents that do. In the same week, a bank and a certifying body drew the human line back. Each item graded for provenance.

A summer watching AI in the art market: record money the auction houses banked without shipping one tool, and 'certainty' sold from products that don't exist.

The art market's loudest AI claim drew its first scrutiny from crypto, not the trade — while the only AI that shipped was quiet throughput below the majors.

You can grant an agent authority over your trades, not over the deciding. On eToro's autonomous connectors, and the question no scope contains.

Weekly AI-in-finance deployment digest: the family-office agentic launch that recurred all summer does not come this week — the run records the absence itself — and in its place ships Astraeus, a wealth-data ontology, the "data architecture is the bottleneck" thesis now productized as a discrete layer. The launches that did land sit at the two ends of the line: eToro wires official Grok and Claude connectors and moves execution to "authorize once, execute within scope" at the