01 / Context
Lots of ideas, too much governance, no system
The global operations function of a banking group operating in 38 markets had more AI and automation ideas than it could process, and no structured route from idea to investment decision. Governance overhead stalled ideas regardless of merit.
Real operational pain sat behind the queue. In payment investigations alone, half a million cases a month required human intervention across the group, at roughly six minutes per case for the operators handling them and three for the checkers behind them.
02 / The Challenge
The fix could not be more governance
The job was to turn an unmanaged flow of AI ideas into defensible investment decisions, without adding yet another layer of governance to the process that was already the bottleneck.
The cost of doing nothing ran both ways: ideas with merit died in governance, and delivery effort went into ideas that were never feasible, desirable, or viable.
03 / The Approach
Hypotheses, experiments, and explicit kill decisions
We designed and ran an innovation-lab front door: a hypothesis-driven experimentation loop, from ideate and assess through business prototype, experiment, and test, ending in an explicit decision to kill, pivot, or persevere. Every idea was assessed on three questions, do they want this, can we do this, should we do this, with AI-specific tests under each: data readiness, model training, user trust in AI outcomes, return on investment, and scaling economics.
Rejection was a first-class outcome with written reasoning: a plausible test-automation case was killed because defect risk outweighed the efficiency benefit, while payment-investigation and repairs-and-exceptions cases proceeded. The loop was paired with customer research: pains, gains, and jobs-to-be-done analysis, and opportunity-solution trees with experiments designed per solution.
And it was built as an operating model, not a personal habit: defined roles, including an AI governance and compliance function, and a ceremony cadence culminating in a monthly deciding forum where stakeholders made the pivot, persevere, or kill calls. The model was adopted and run in practice, not left as a proposal.