Feb 2026
We tried everything. Some tools stayed, many were binned. Here's how we use AI at the agency without losing the human touch that sets us apart.
Every week a new AI tool comes out promising to revolutionise design. The reality is that most don't survive more than two months in our stack. Ludmi tests practically everything that appears, and she has a pretty accurate nose for separating the useful from the hype. But some tools did stay and changed how we work.
The first to earn its place was for initial copywriting drafts. When we're building a website, we need text that makes sense so we can design with real context. Before, Ludmi used Lorem Ipsum and the client couldn't understand the proposal at all. Now we generate contextualised copy drafts in seconds, which allows us to present prototypes that already feel real. Mind you: I then review everything and inject our tone: AI provides the skeleton, we add the personality.
Another surprise: AI for verifying contrast and accessibility. Juli uses it a lot when working on colour palettes for branding. Before it was a tedious manual process; now we have tools that analyse palettes and instantly tell us if they meet WCAG AA. It's saved us more than one post-launch headache.
For automations, we use AI integrated with our workflows. For instance, when a client completes the contact form, a system classifies the project type, estimates complexity, and assigns it to me internally. The client receives a personalised response in minutes, not hours. I built that because I got tired of triaging emails manually at 11pm.
Where we're most cautious is in pure visual design. Michael tested UI generators and the team's conclusion was clear: they're useful for exploring ideas quickly, but the final result always needs Ludmi's or Juli's touch. Good design isn't just about looking pretty: it has to respond to business objectives, and that still requires human judgement. Even Susi tried AI video editing tools and reached a similar conclusion: they speed up mechanical work but she still provides the narrative.
The advice we give: don't try to use AI for everything. Identify the bottlenecks in your process, find specific tools for those pain points, and measure whether they actually save time. In our case, we estimate it saves us about 8 hours per week. That's an entire day we dedicate to thinking better, not working more.