An AI-native product team
Moved a product team of around thirteen to directing coding agents in parallel, with sandboxed environments, role-specific instructions and multi-stage skill pipelines.
About
Tom has spent 18 years optimising how software gets built in regulated environments: investments, pensions, payroll, advice, crowdfunding and identity, at companies including Seccl, Redington and Altus/Equisoft.
Most recently, as CTO and co-founder of an FCA-regulated UK fintech, he led a team of engineers, product owners, designers and subject-matter experts who directed coding agents as a normal part of every day, and built much of the tooling, process and governance behind it.
dulcedo brings that experience to other teams, so they can make the same shift safely.
I’ve been running a team where agents do most of the typing and people make the calls. It works, but only when the rules, guardrails and checkpoints get as much care as the code. I started dulcedo to help other teams get there without learning it the hard way.

A few things Tom has designed and run that shape how we work with clients.
Moved a product team of around thirteen to directing coding agents in parallel, with sandboxed environments, role-specific instructions and multi-stage skill pipelines.
Reworked ceremonies so people agree intent up front, rather than spending hours in meetings reviewing what agents produced.
A bespoke workspace that shows every micro-factory at once, so people step in only where they’re needed.
An engineering levelling framework with agent direction and leverage as a core dimension. Judgement is the key skill, not execution.
Merge queues, affected-test selection and smoke-test gating to keep the pipeline ahead of a rising pull request volume.
Designed cloud workspaces for agent-driven development, with single sign-on, idle shutdown and a per-engineer cost model.
Thirty minutes to talk through where your team is today, what’s slowing it down, and whether we’re the right people to help. If we’re not, we’ll say so.