Steady standards in a fast world
Estimated read time: 6 minutes
Speed may be the headline of a vibe-coded world, but trustworthiness remains the bottom line. When we can deploy changes with unprecedented pace, our standards and guardrails matter more than ever. If you’re going to drive faster, your brakes and seatbelts need to be reliable. We must uphold, and in places strengthen, our quality benchmarks, so velocity yields faster value, not faster failures.
The UK Government’s Service Standard (alongside equivalent guidelines in other countries) is the foundation here. If delivery accelerates, its criteria don’t get diluted just because we can move quickly. On the contrary, they become load-bearing. Any adoption of AI-assisted delivery should happen within the Service Standard, not around it or in defiance of it.
And if that’s the bar, then the next question is practical: where does that bar live, day to day, when more people can ship?
Standards have to live in our defaults
A vibe-coded world changes where standards have to live. When more people are getting their hands dirty in the codebase, then we need to find ways to ensure quality is baked into the defaults of our tools, including the prompts, policies, and guardrails we give to AI agents, as well as the working practices of the team itself.
So even if we can now spin up a working service in a day, we still meet the basics before it is in front of real users. If AI helps to write code, it still goes through review, testing, and continuous integration. Accessibility, security, data protection, and operational readiness don’t become optional because the tool is quick. In fact, they become more central, because you can now change more, more often.
The good news is that AI can strengthen the basics too: generating test cases, improving monitoring, and helping teams spot issues earlier. But none of that removes judgement. It just makes it harder to justify not doing what we already know protects our users.
Make the safe path the fast path
If we want standards to hold at higher speed, we have to stop treating safety as an extra step. Modern platform infrastructure can make the safe path the fast path: secure, scalable environments with the right defaults baked in; one-click deploys; monitoring and logging; feature flags; and easy rollback.
We don’t want a world where a team can build an MVP in a day, but then spend three months waiting for security clearance or someone to provision an environment. Keeping standards steady is not only a team discipline. It is an organisational one. It means investing in enabling systems – platforms, pipelines, automated testing, and operational tooling – so that the path of least resistance is also the right one.
In many public sector organisations, the bottleneck to rapid iteration isn’t writing the code. It’s the surrounding bureaucracy and legacy infrastructure. Overcoming that is as much culture and leadership as it is tech: insisting that governance and delivery pipelines adapt, and making it easy to deliver secure, accessible, documented and supportable change.
“Done” does not get cheaper just because “build” does
A fast world tests discipline. There will be temptations to skip steps: “It’s just a trial, can we bypass some tests or design reviews to save time?”. The discipline is to resist. If we use AI to generate code, we can also generate tests and run them immediately. If we release to a small percentage of users, we put analytics and feedback channels in place from day one. Our definition of “done” does not soften because “done” arrives sooner. Every change still has to meet the bar: does it work for users, is it accessible, and does it operate as well on a bad day as a good one?
The most dangerous moment is often the one where it looks finished to a busy senior person, while the team can still see all the ways it isn’t ready to carry the public.
And “done” has consequences beyond the deployment. It has a human dimension. AI can make delivery faster; it cannot make people inexhaustible. If speed is coming from heroics and burnout, then we haven’t become more capable. We’ve just shifted the cost onto individuals, and it will surface in the health of the team and the reliability of the service.
“Done” also has a footprint as well as a finish. AI-assisted delivery leans on compute, and compute has real costs: financial, environmental, and sometimes geopolitical. Stewardship means we don’t treat that as someone else’s problem. Be deliberate about where we use these tools, avoid wasteful patterns just because iteration is easy, and make it normal to ask: what are we spending – in money, energy and dependency – to get this result?
Trust is what people take away
Faster does not mean sloppy. Faster means more value delivered so long as quality is maintained. If speed comes at the cost of excluding users, exposing data, or confusing people, it’s not progress at all. So our job is to ensure that doesn’t happen.
Because the broader implication of steady standards is public trust. Citizens may not know anything about vibe coding or agile or AI, but they certainly experience the outcomes. If government services start improving visibly week by week, that can build confidence in public institutions’ responsiveness. But if things break faster, or change unpredictably without proper support, then trust erodes.
Keeping standards means keeping the implicit social contract: we move quickly and fix things. We innovate and uphold reliability. That balance reassures the public, and those who oversee our work, that we aren’t adopting new tools as a reckless gamble but as a thoughtful evolution.
Faster practices demand a level head and a steady hand. By holding firm to service standards and our ethics, we can ensure increased velocity translates into increased public value, not just activity. We haven’t taken our eyes off the destination: services that are accessible, secure, accountable, and centred on the people who need them. We just have an opportunity to reach that destination sooner.