Kai-Fu Lee wrote a piece in the Financial Times yesterday about what happens to middle management once AI agents start doing the work. He was writing about corporate hierarchy, but I read it as a problem every exploration or mining finance team will recognise.

For context, Lee set up Microsoft Research Asia, ran Google China and now runs 01.AI. His new book is called AI Native.

His argument is that a company is an information system, and most of what it knows sits in "dormant data": documents, records and conversations it produces and never uses. He describes an internal system that found two teams who had spent three months building the same thing, repeating each other's mistakes, without anyone noticing.

Most mining companies I speak to have a version of that. Drilling from three previous owners in three formats. Assays in a consultant's spreadsheet. A 1990s technical report explaining why the last operator walked away, scanned, in a language half the team can't read. Multiple sources, no common structure, and no reliable way to know what you don't have.

At a talk earlier this year, Lee blamed most corporate AI failures on missing context: garbage in, garbage out. His fix starts with an ontology, a structure the AI can understand, before switching on any agent. That is what we are building with the Pulse Data Core: a company's own geological, operational and legal data, whatever the source, brought into one system of record that AI can work across, with every figure traced back to the page it came from.

Vlad Tiazhov, who leads our engineering, put the consequence bluntly to me recently. "Capacity is becoming a commodity. Judgement is where the value sits now".

For a geology team, that is good news. An agent can read every drill log in the archive. It cannot decide where the next hole goes, and it cannot sign off a resource as a Competent or Qualified Person. Lee says AI cannot bear responsibility. In mining, that responsibility carries a name, by regulation.

Quote card: 'What AI cannot do is bear responsibility.' Kai-Fu Lee, Founder and CEO, 01.AI. In mining that responsibility carries a name: the Competent Person signs the resource.

He also warns that moving too fast erases institutional knowledge. The aim should be the reverse: get what the senior geologist knows, and what is buried in the archive, into a form the whole team and its AI can use, so the people with the judgement spend their time using it.

Less searching. More strategising.™

AI Readiness Diagnostic

Where does your team's data infrastructure sit today?

Answer 10 questions. Get a private diagnostic on your AI readiness, in minutes.

Pulse Intelligence

Less Searching. More Strategising.™

See the platform running on real mining data, and what it would look like for your team.