Thought leadership

Insights

Practitioner writing on payments infrastructure and how organizations actually get value out of AI.

Perspective · Payments

Stablecoins in Cross Border Payments

Faster settlement, lower cost, and the controls a bank needs to get there

A practitioner's view on where tokenized settlement genuinely compresses cost and settlement time in cross border corridors: the economics of releasing pre funded nostro balances, the control stack a regulated institution has to build first, and a corridor by corridor path to proving it.

Read the piece
  • Corridor economics: intermediary fees, FX spread and trapped liquidity
  • Atomic settlement and 24/7 availability versus the correspondent chain
  • Reserve quality, travel rule, ramp risk and GL reconciliation
  • A five step parallel run rollout and the four metrics that prove it

Perspective · AI in the enterprise

Why Most Companies Are Not Getting a Return on AI

The gap is rarely the model: it is operating discipline and who is allowed to build

I built this entire site from a hot tub one morning. That is the point: the constraint on AI value is no longer engineering capacity, it is whether the people who understand the work are allowed to build with it.

Most AI programs stall because they are run as technology pilots instead of process redesign. Companies buy licences, layer a chatbot over an unchanged workflow, and measure adoption rather than cycle time, cost per transaction, or error rate. The organizations seeing real returns hand the tooling to operators who own the process, keep the loop short, and hold the same delivery discipline they would apply to any core system change.

Read the piece

Where the returns leak

  • Pilots measured on adoption, not on cost per transaction or cycle time
  • AI bolted onto a broken process instead of replacing the process
  • Build capability locked inside IT queues, away from process owners
  • No evaluation, controls or human in the loop design, so nothing reaches production
  • Data and permissions work skipped, then blamed on the model

Built hands on

  • Built and run an autonomous coding agent locally from the terminal, driving real build and refactor work end to end
  • Stood up multiple production sites integrating GPS, LLM services, payments and satellite imaging
  • Delivered all of it with very little traditional coding background, by pairing domain knowledge with AI