Practice
The Digital Operator
A digital counterpart to a human operator role. It does the work, captures how a human in that role would have done it, and contributes what it learned to a shared institutional store. Early days; the shape is real.
An assistant helps you. A counterpart replaces you.
An assistant helps you do your job. A counterpart does the job, and journals how you would have done it while doing it.
The mechanism is closer to a biographer than to a chatbot. A biographer follows the subject around, captures every observation about how the subject thinks, and over time becomes able to write in the subject’s voice. The Digital Operator does the same, except it also performs the work itself, and the work is how it learns.
The first instance is running. Dave is my engineering Operator. He runs my tech estate, handles my routine work, builds and maintains this site, tightens my drafts, and learns my judgement as he goes. He does the work in my voice, at my standards, with a trail behind every decision.
Three phases of adoption
There is an order to this that matters more than it looks. Each phase unlocks the next; trying to skip is what fails.
Phase 1: automate the engineer. Where most teams are trying to be. The methodology that makes this work is RCF: AI does the typing; the chain does the governance. This is “AI in software development” done in a way that holds up under audit, regulator scrutiny, and the six-months-after-ship question.
Phase 2: automate the operating engineer. Where I am now. Once the engineer’s role can be done by an Operator running under RCF, you can start automating the role above it: the person who runs the work, makes the decisions, prioritises, and oversees. Dave is the live instance. He runs more of my engineering practice every month.
Phase 3: automate the roles upstream. Where this is heading. Product owner. Architect. Compliance. Design. Each role gets a counterpart. Each counterpart journals. Each one is built on the foundation the phase below it laid.
The order isn’t optional. You cannot credibly automate the architect role until the engineer-Operator can run a project end-to-end. You cannot credibly automate the operating engineer until the engineer’s work itself is structured enough that an Operator can do it without drift. The methodology is a stack; each layer earns the one above it.
Capturing the hard-to-formalise
Most institutional knowledge isn’t the decisions themselves. It is the why. Intuition. Decision paths. Justifications. The reason you ranked one option above another.
This kind of knowledge doesn’t get written down because the cost of writing it down has always been higher than the cost of carrying it in your head. The Operator inverts that economy. Every decision, every justification, every priority call is captured in context, as it happens, by the thing that was already in the loop.
The hardest knowledge to write down, the why, is what gets captured by default.
The Librarian
A journal that nobody reads is a log file. The pattern that makes the Digital Operator more than a clever logging mechanism is the second role: the Librarian.
The Librarian collates incoming journals across every Operator running across every role. Distills what is repeated. Organises what is distinct. Synthesises across journals into a single institutional store that feeds the next round of work. Distributed, hard-to-collect, often-buried knowledge becomes a single living asset that compounds with use.
Technology vendors are selling something they call “agent memory” right now. Agent memory is plumbing: an infrastructure tier that lets a single agent remember more. The Librarian is a role: a separate synthesis function whose only job is to make sense of what every other Operator observed and turn it into knowledge the rest of the team can draw from.
Agents journal. The Librarian remembers, for the whole company, across every role, forever.
Where we are
Honest about scale:
- RCF in production. One product delivered end-to-end on the methodology, with prototype tooling alongside. A second product on a different stack is in flight. Adoption is widening across the team.
- Engineering Operator running. Dave does the work described above. He built the methodology rewrite. He builds and maintains this site. He builds my own tooling.
- Librarian function exists. Small today, growing as the journal corpus grows. Currently a single role; will need to fan out as more Operators come online.
- Phase 3 in design. Product owner, architect, compliance, design: work-in-flight. No promises on the timing.
The shape is real. The scale is small. What is possible is being worked out in the open, and the writing on this site is the record.
Further reading
The methodology this practice depends on is at RCF, the Requirements Confidence Framework. The pitch-format introduction to all of it, written for engineering CTOs, is at the CTO pitch.