Artificial intelligence has become part of how a lot of software teams work, and ours is no exception. Rather than be vague about that, we would rather be plain: here is where AI helps us, where it does not, and why the way we work has not really changed underneath.
The short version is that AI is a useful tool with clear limits. It is good at getting a first draft onto the page, at helping us read around a problem quickly, and at producing the repetitive scaffolding that every project needs. It is not good at deciding what is correct, what is safe, or what is right for a particular client. Those decisions still belong to a person, and on our work they always will.
What this adds up to for a client is straightforward. You get the speed that these tools genuinely offer, without inheriting the risk that comes from trusting them too far.
What we actually use it for
Most of our day-to-day use of AI falls into three areas. The first is drafting. A function, a configuration file, a first pass at documentation, an outline for an article like this one: getting from a blank page to something we can react to is often the slow part, and AI shortens it. Reacting to a draft is faster than producing one, and the draft is a starting point, never the finished work.
The second is research. When we are weighing an approach, checking how a particular Drupal interface behaves, or getting oriented in an unfamiliar corner of a codebase, AI can gather the relevant background quickly and point us at the right documentation to confirm. We treat what it says as a lead to verify, not an answer to trust, because these tools can be confidently wrong.
The third is boilerplate. Every project has repetitive, predictable code: the standard shape of a module, a set of similar test cases, a data structure that follows an established pattern. This is exactly the kind of work where a tool that has seen the pattern many times can save real time, and where a person can quickly tell whether the result is right. The time saved on the routine parts is time we can spend on the parts that are not routine, which is where it belongs.
Across all three, the shape is the same. AI does the part that is mechanical or fast to check, and a person does the part that needs understanding. It never gets the last word, and it never removes a step; it just makes some of the steps quicker to reach.
What we do not use it for
We do not let AI make the decisions that matter. It does not choose your architecture, decide how your data is modeled, or determine how something sensitive is handled. It does not sign off on security. It does not decide when something is finished. Those judgments depend on understanding your business, the trade-offs involved, and the consequences of getting them wrong, and that understanding is ours to hold.
We are also careful about what these tools ever see. Client data, credentials, and anything private stay out of them. The help we take is with the shape of the code and the thinking around it, not with anything confidential entrusted to us.
Every change passes through the same gate
Here is the part that matters most. Whatever helped produce a change, it goes through exactly the same review as anything else we write. A person reads it, line by line, and is responsible for it. It has to meet the same coding standards, follow the same Drupal conventions, and fit the patterns already in your codebase. It has to pass the same automated tests before it can be deployed, and it goes out through the same tracked, reversible release process we use for every change.
Nothing gets a shortcut because a tool was involved. A draft that AI helped write and a draft a developer wrote by hand meet the identical bar before either becomes part of your site. From the outside, and in the code itself, you cannot tell which is which, because by the time it ships it has been shaped and checked to the same standard either way.
Why the review is the point, not a formality
It would be easy to treat that review as a rubber stamp, and that is precisely the mistake we avoid. These tools produce work that looks plausible, which is exactly what makes careful review necessary rather than optional. Code that reads well can still be subtly wrong, miss an edge case, or solve the wrong problem cleanly. Catching that takes someone who understands the whole system and is accountable for the result.
So the senior review is not there to rescue the occasional mistake. It is the step where the actual engineering judgment happens: does this belong here, does it hold up under the cases that matter, will it still make sense to the next person who reads it in two years. A tool can accelerate the typing. It cannot take on the responsibility, and we do not ask it to.
What this means for your site
The practical effect is that some work reaches you faster than it would have a few years ago, and the parts of a project that used to be slow for no good reason, the scaffolding and the first drafts, take less time. That leaves more of our attention for the parts that genuinely need it: the design decisions, the tricky logic, the careful review.
What does not change is the standard the finished work is held to. The code on your site, whether a new feature or ongoing maintenance, is code a person understood, reviewed, and stands behind. You get the benefit of the speed, and we keep the responsibility for the quality. That balance, using new tools without lowering the bar, is the whole of our position, and we think it is the honest one.
Where we think this is heading
These tools will keep improving, and we will keep adjusting how we use them as they do. We are neither breathless about AI nor afraid of it. It is a capable assistant that has made parts of our work quicker, and like any tool it is only as good as the judgment applied around it.
Our commitment is simply this: however our tools change, the work we hand you will always have been reviewed, tested, and understood by a person who is accountable for it. That has been true since before any of this arrived, and it will stay true. If anything, the more these tools can do, the more the human judgment around them is what you are really paying for.