In December 2025, Born Digital began to train its team in agentic development, gave them firm-provided AI tooling, and rewrote its delivery workflow around it. The aim is simple: to use new technology as soon as it is dependable, and to use it under rules that protect our clients.
What agentic development means here
Agentic development means an engineer directs AI tooling to carry out parts of the work, and then checks the result, rather than writing every line by hand. The engineer still decides what is built and how. The tooling reduces the routine work it takes to get there.
Why we adopted it now
By late 2025, AI tooling had become capable of reliably speeding up a specific kind of engineering work: work that is large, repetitive and easy to check. The tools were going to arrive in every development team regardless. We chose to adopt them deliberately, with shared training, tools provided by the firm and clear rules, so that the standard does not vary from one engineer to the next.
Where it helps, and where it does not
The tooling is faster for migrations, test coverage and refactoring. Moving a site's content and templates to a new platform, for example, involves a great deal of repetitive work that can be checked line by line. That is the kind of task this tooling suits.
It is no help for judgement. Deciding what a homepage should say, what a business should prioritise, or what to build in the first place remains work for people, and we treat it that way.
The rules we work to
- Every AI tool the team uses is provided and paid for by the firm. Nobody works on client projects through a personal account.
- Client credentials are never given to AI tools. Client content and data are only used with tools that do not train on them.
- Decisions about what to build, what to say and what to prioritise are made by people. The tooling carries out production work under their direction.
- No AI output reaches a client's live site without a person approving it. The engineer who approves a change is accountable for it.
The risk we guard against
Faster production has a known failure mode. It makes it easier to produce a large amount of mediocre work quickly, and a team that stops reading its own output carefully will ship worse software than before. The approval step exists for exactly that reason, and it is the part of the workflow we are least willing to relax.
What it means for cost
Work that is large, repetitive and checkable now takes less engineering time than it did. Engineering time is a large part of the cost of any technology project, and reducing it is how we keep the solutions we offer cost-effective for our clients.
To find out how this applies to your own project, get in touch with us.