If you didn't need an assistant before, why do you need an AI assistant now?
That's where most teams get stuck. They hear "AI" and immediately go into transformation mode, ripping apart workflows, trialling seventeen tools, trying to automate everything at once. The result? A very expensive mess that runs slower than the spreadsheet it replaced.
Here's the uncomfortable truth: not everything should be touched by AI.
Some things need a human. Some things belong to your existing systems. And some things, the repetitive, manual, mind-numbing stuff, are genuinely where AI earns its place.
So before you sign up for another tool, do something boring first. Look at your actual process. Not the idealised version of it. The real one. Ask yourself...
- Where is your team spending hours on work that follows the same pattern every single time?
- Where are things being copy-pasted, reformatted, chased, re-entered?
That's your starting point.
Pick one of those areas. Automate that. Nothing else.
Here's what that looks like in practice.
For one of our clients, designers and BA's were spending significant time manually checking whether every use case had been designed and had all the negative states been covered. The empty states. The edge cases everyone forgets until QA.
It was important work, but it was also slow, repetitive, and depressingly easy to get wrong.
We didn't rebuild their design workflow. We took three inputs, requirements, acceptance criteria, and the Figma file, fed them into Claude, and built a plugin that does the validation automatically. That's it. One small, specific piece of the process.
The outcome
~20% of designer and BA effort saved. Not from a sweeping transformation. From fixing one bottleneck.
The trap isn't failing to adopt AI, it's adopting too much of it, too fast, in too many directions. You end up with a collection of clever tools that don't talk to each other, and a team that's somehow doing more manual work than before, just in different places.
The skill that actually matters right now isn't finding the best AI tool. It's knowing how to stitch your processes together after you've automated a specific part. That's what makes the whole thing compound. That's what makes it worth it.
Start narrow. Stay with one tool long enough to make it work. Then build from there.
AI isn't going to save your delivery. But used in the right place, in the right way, it'll give you back the time to do that yourself.