Speed without structure and direction is like letting the air our of an inflated balloon. It makes fast progress but soon loses direction with unpredictable results and you hurt yourself knocking over a chair trying to jump out of it’s way.
This strained metaphor is about trying to get sustained speed from Gen AI design tools. Ok let’s try again…
A well run design function is sustainably fast with or without AI design tools. That speed comes from good Design Ops and following good design processes. With good processes and a working design system the gap between ‘confidently understanding the requirements’ and ‘ready to build’ approaches zero.
Mini-case study of design systems creating speed
Design systems let you design without designing. At Thread, during our Beta phase we were still adding features. One feature that needed adding was the Conditions page. Thread is/was a utility asset inspection SaaS product where asset managers and others could view the state of all their assets. We started with the primary asset centric view of the information users to explore their assets. Each asset could have multiple ‘conditions’, things that are wrong with it. This was a powerful default, but soon users wanted to explore the informatoin in from condition centric view of the information – “Show me all the severity 3+ conditions on insulators reported in the last 6 months”. This needed a relatively complex new page.
Thankfully we had design system. In the design system we had:
- A data exploration page template
- A data table component
- All the possible table cells and how they supported different types of data
- Filters
- Action bar
- And more
Everything needed to build the page.
The design process involved adding a link to the top level navigation, then working with the product team to define the data that was needed in the table. That’s it. No high fidelity UI design. Not even a need for wireframes. This was design by specification, the requirements specced the information and behaviours, the design system specced the UI and interactions.
The design time could not have been much faster.
Strong foundations + continuous improvement = useful speed
AI design tools are often premised on speed. But design speed isn’t usually the bottleneck in a delivery process. If design is the bottleneck then improving your design processes should be the first response, not accelerating and amplifying poor processes. Don’t just let go of the balloon and hope for velocity, even if it does feel fun.
The more you invest in good design practices and in particular design systems the more the time to deliver UI designs approaches zero.
Good process improvment isn’t exactly new. Each week, each sprint, each release, ask what’s the biggest bottleneck we have in delivering designs fast effectively and focus on improving that. Sometimes that will be best served by an AI tool, sometimes it won’t. And sometimes an AI tool can let you build a non-ai tool that improves that bottleneck. AI gives us a new repertoire of tools and abilities but the important thing raimains improving the design process – not using AI for the sake of AI. Design leaders should have a design delivery strategy, not an design AI strategy.
Identify and fix your bottlenecks one at a time, use AI tools where they are useful.