Lloyds says payment agents stay small and human-supervised
A major bank's actual pattern: AI as reasoning engine, deterministic workflows for execution, humans in the loop — autonomy remains distant.
Original event 2026-09-25
Clare Schramm, Managing Director of Transaction Banking Technology Platforms at Lloyds Banking Group, told diginomica that the bank is trialing agents in its payment journeys, but the use cases are small and always supervised by humans.
The pattern: generative AI reasons and problem-solves, then hands off to a deterministic workflow for execution — for example, proactively correcting address errors in payments before they go through, freeing staff for higher-value work. In foreign exchange, AI predicts favorable rates for customers; Schramm says there has been real bottom-line impact, without giving numbers.
She stressed that AI at the bank is "still very much in research and development mode" and that customer trust demands heavy guardrails. The bank uses Microsoft Copilot, Anthropic's Claude Code and Google Gemini. She also advocates open standards for agentic payments, convened through the open source community; without a common protocol, she says, the whole agentic promise collapses.