Day 2 · Recorded 7 May 2026

How Monzo enables AI tools without losing control

For regulated teams, the answer cannot be "let everyone paste data anywhere" or "ban the tools and hope." Suhail Patel shows how Monzo's opinionated platform, 3,000+ microservices in a monorepo, static analysis, and data-flow controls make AI adoption more practical precisely because the engineering system already has strong defaults.

Suhail Patel, Principal Engineer, Monzo

What's in this session

Monzo's AI advantage starts before AI. A decade of golden paths, service generators, standardized libraries, and static analysis gives LLMs a codebase with recognizable structure, while giving the platform team real control over how production systems are changed.

In this interview, Suhail Patel explains how Monzo rolls out AI tools inside a regulated bank: opening access, setting clear rules for data flow, sandboxing untrusted work, governing MCP access, and owning the interface to models rather than tying the company to one vendor shell.

For engineering leaders, the session is a pragmatic alternative to blanket restriction. Useful AI adoption depends on meeting engineers where they already work while making the secure path the easy path: observable, controlled, and compatible with the standards that keep the bank running.

Inside the recording

  1. 00:00 Monzo's platform advantage
    A regulated bank with thousands of services starts from strong golden paths and internal tooling.
  2. 04:00 AI as a new engineering muscle
    Suhail explains why judgment, review, and asking the right questions matter more than ever.
  3. 16:00 Static analysis and code mods get cheaper
    AI lowers the cost of writing checks, migrations, and rewrite tools for a large Go codebase.
  4. 20:00 Background agents without losing monorepo control
    Greenfield freedom does not translate directly to production systems with shared standards.
  5. 28:00 Controlled access beats blanket restriction
    In a bank, useful AI adoption depends on clear data flow rules, sandboxing, and governed usage.
  6. 32:00 Owning the interface to AI tooling
    Why Monzo wants optionality across models and tools instead of coupling itself to one vendor shell.