Sunday, 4 October 202602:45 UTCWire updated 02:45 UTC

Follow @0xNotMarc

Notes from the Terminal
Delayed ·

The Morning Note

Finance is hiring for agents before it can fully govern them

Wall Street's AI-agent hiring boom arrived alongside new agent deployments in insurance and payments, a major lab's training push, and a fraud case showing the same technology cuts both ways.

The hiring signal

Postings for AI-related roles at banks including JPMorgan Chase, Citigroup and Capital One rose 49% this year to 139,819 listings, and postings mentioning agent orchestration jumped 1,721%, Draup found. Draup's own CEO calls it arguably the hottest skill on Wall Street, and generative AI managers already earn a median base salary of about $190,000.

Agents already doing the work

Agents are already doing that work. Liberate's AI agents answer insurance policyholder calls and log them in a carrier's core system, and Liberate says they have given insurers and brokers back more than 100 million minutes. More than 70 carriers and brokers run over 3.5 million transactions a month on its platform, and Branch, a home and auto insurer, cut claim reporting time 42% after moving first notice of loss to Liberate's agents.

Payment rails are catching up

Card networks are trying to keep pace with what agents are now allowed to do. EMVCo closed public comment on Sept. 30 on a framework for Intent Services that would let payment participants register what a consumer authorized an AI agent to do, before, during and after a transaction, and it may add know your agent checks to identify agents in a payment. PYMNTS Intelligence found 56% of consumers would let an agent search and compare products, but just 37% would let one authorize a payment.

Labs are training the supervisors

Anthropic is investing $100 million in a Claude Frontier Academy that aims to train 10,000 frontier deployed engineers by the end of 2027, with its first groups including engineers from Accenture, Morgan Stanley and Novo Nordisk. Anthropic's Steve Corfield says the academy trains people the way its own engineers learn.

The same capability cuts both ways

The same voice-cloning capability that powers legitimate agents is already being turned against people directly. In Malang, a scam cloned the voice of a victim's close contact with AI and used it to talk its way into a transfer of Rp80 juta, OJK's local branch disclosed.

What to watch

Put together, this is one week where agents moved from answering questions to handling calls, payments and transfers on their own, which is exactly why the hiring, the standards and the fraud are all accelerating together. For a bank or insurer, the useful question is whether the people, the payment rails and the fraud controls built around these agents are growing as fast as the agents themselves.

The stories behind this note

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5