We use AI tools daily across design, development, and client communication. Some of it has become simply how we work now. Some of it we have tried and quietly stopped using. Here is the honest version of both.
Where AI genuinely saves us time
First-draft copy.
AI-drafted first passes on page copy, meta descriptions, and email replies give us a starting point in seconds instead of a blank page, which we then edit for accuracy and voice.
Code assistance.
AI-assisted coding speeds up repetitive front-end work such as building out sections and components, particularly for patterns we have built many times before.
Summarising and documenting.
Turning a messy client call or a long email thread into a clear brief or set of build notes is one of the most reliably useful things AI does for us.
Where it falls short
Anything requiring real client context.
AI does not know a client’s history, their previous frustrations, or the unwritten reasons a certain approach was already rejected. That judgement still has to come from the person on the account.
Final design decisions.
AI-generated design suggestions are a useful starting point but rarely reflect the brand nuance and taste that make a design actually fit a specific business.
Anything security or infrastructure related.
We do not let AI make unsupervised changes to live client sites, server configurations, or security settings. The risk of a confident but wrong suggestion is too high in this area.
The rule we work by
AI drafts, a person reviews. Every AI-assisted output that reaches a client has been checked by someone on the team who understands that specific client’s context. That single rule is what keeps AI useful instead of risky in agency work.