AI has changed how we build software. Not in the way the loudest posts claim, but in a steady, practical way that shows up in how quickly we ship and how much we catch before we do. Here is how we actually use it, and where we are careful.
AI does the repetitive work
Most of our engineers use AI every day, for drafting and reviewing code, generating tests, writing documentation and getting to grips with an unfamiliar codebase. It is quickest on the work that used to eat time without needing much thought: boilerplate, the first pass at a test suite, the tedious refactor across dozens of files. That frees people up for the parts that genuinely need judgement, which is where the value in a senior engineer actually sits.
Quality does not move
The bar for what ships does not change because AI helped write it. Everything goes through the same review and the same tests, and a named engineer is accountable for it. We treat what AI produces as a first draft to be checked, not an answer to be trusted. Moving faster is only worth anything if the quality holds, so we keep the checks exactly where they were.
We are careful with your code and data
We do not send client code or data to third-party models without agreement, and we are clear about the tools we use and where they run. A person makes the call on what goes in. AI informs the work. It does not sign it off. If you have rules about where your data can go, those come first.
What the evidence actually says
It is worth being honest about the research, because it is mixed. A randomised controlled trial of GitHub Copilot found developers finished a set task around 55% faster, with the biggest gains for those earlier in their careers. A 2025 study from METR, looking at experienced developers working on code they knew well, found the opposite: they were about 19% slower with AI, even while they felt faster.
Both can be true. AI speeds up some work and gets in the way of other work, and the skill is knowing which is which. That is why we lead with engineering judgement and treat AI as a tool in the hands of people who have done the work before, rather than a shortcut around them. It is also why we will not quote you a headline number we cannot stand behind for your situation.
Helping your team do the same
If you want to use AI in your own delivery, we can help you do it in a way you can trust: working out where it genuinely helps, setting the review and data-handling rules that keep it safe, and upskilling your engineers so the capability stays with you after we leave. There is more on our AI page and our software engineering page.
If you want software delivered well, with AI used where it earns its place, we would love to hear from you.