Citizens Build, Agents Execute, Experts Govern
AI has made software dramatically cheaper to create and left the cost of trusting it almost exactly where it was. This series works through what follows: why the weekend app is real software, what changes the moment a business starts depending on it, and how citizen builders, AI agents and expert engineers work together inside guardrails that scale.
10 articles
Start here
The gap between creating software and being able to depend on it, and why it did not close.
The three roles
What business users bring, what AI agents genuinely do, and why engineering judgement gets scarcer.
- Part 2Citizens build: why business users should be part of software creationThe procurement officer who builds a tender checklist assistant knows something no developer does. The prototype is rarely worth keeping — but the requirement buried inside it usually is.7 min read
- Part 3Agents execute: what AI agents actually do in software deliveryAn agent can read a request, find the right files, write the code, add the tests and open a pull request. What it cannot do is decide the change was a good idea — and that is the step the pipeline has to enforce.8 min read
- Part 4Experts govern: why engineering judgement becomes more valuable in the age of AIWhen generating code becomes cheap, the expensive part is deciding which code should exist. The senior engineering role is not shrinking — it is losing the half that was mechanical.8 min read
The operating model
Putting the three together — and the anti-pattern that appears when the order is wrong.
- Part 5The new software factory: citizens build, agents execute, experts governThree roles, and an order they have to happen in. Run the same three steps in the wrong sequence and you get the anti-pattern instead of the operating model.9 min read
- Part 6The AI delivery anti-pattern: business builds, engineering cleans upNobody designs this arrangement. It emerges when a tool becomes load-bearing without anyone deciding it should — and by the time engineering is asked to make it production ready, the expensive choices have already been made.8 min read
Making it practical
A checklist for promoting a prototype, and the platform work that lets governance scale.
- Part 7From prototype to production: the trust checklist for AI-built softwareEight questions to ask before a business depends on something that was built quickly. None of them are about whether the code works — that part was never in doubt.8 min read
- Part 8Platform engineering for the citizen-and-agent eraA rule is followed when somebody remembers. A default is followed when nobody does. When the number of builders multiplies, that difference stops being a preference and becomes the whole strategy.8 min read
In the real world
What the model means for public service delivery, and where CloudNala starts.
- Part 9Public-sector AI: citizens build, agents execute, experts governA chatbot added to a workflow that cannot answer simply gives citizens a faster route to the same silence. In public service delivery, accountability is the design constraint — not an afterthought.8 min read
- Part 10CloudNala's view: governed AI delivery for real business workflowsAlmost nobody starts at step one. Most organisations arrive holding something that already works and now matters — which makes going back and doing the foundations feel like going backwards, right up until it doesn't.7 min read
People are already building. Do you know what exists?
Most organisations discover their AI-built tooling one incident at a time. A short discovery and triage pass is usually the cheapest way to find out what is running, what data is moving, and which two or three tools have quietly become load-bearing.
Book an AI Readiness Workshop