It’s the end of design as we know it, and I feel fine…
Design is changing fast. AI tools such as LLMs, Agents, and other automation tools perform some tasks remarkably well. As such, Designers can be empowered to execute parts of the workflow with greater speed, fluency, and accuracy.
Here’s a summary of some areas where I’m seeing AI impact emerge, with more info in the coming weeks…
Research
User Research is critically important, however it is often believed to be prohibitively labor intensive
- Crunching large datasets to find potential interview participants
- Intelligent and automated scheduling of sessions
- Capturing, transcribing, and summarizing interviews
- Finding similarities and trends across interview data
…and so on. Not perfect. Not suitable for every occasion… but making inroads daily.
What is emerging in the development and usage of AI tools is the importance of Context Engineering. Effective user research is a critical input into defining the context of a given product and will likely pay an outsized role in shaping Context Input (for organizations that can embrace it).
Ideation
Teams often need hours* to move beyond obvious ideas. AI is not replacing human brainstorms, but it can serve as a partner:
Research has shown that, in brainstorming situations, people benefit from taking a pretty long amount of time* to begin gnerating ideas that are truly effective.
I don’t have the confidence (or naiveté) to believe that AI is replacing human ingenuity en masse.
Tools like a GPT can be put to use in the same way that a screenwriter will lean on a writing partner.
- Write an idea
- Ask the partner to come up its take on the same idea
- Compare the two
- Refine the original
…and so on.
An LLM’s ability to produce seemingly infinite quantity of ideas is often mistaken for its ability to produce those ideas at a level of quality that is both innovative and valuable.
Ideation sessions benefit from a kind of decomposition and reconstitution that AI does well. An LLM will examine a given input without the context (or baggage) that a team may have. It’s a mechanical approach to the task, leaving the judgement and evaluation to the humans in the loop.
I spent time with the team at McKinsey working on this problem in a business innovation context and found it fascinating. Experts in a give field leveraged AI to generate new innovations and mine weaker outputs for salvageable parts. The team weighed the options presented, leading to a secondary conversation determining if any of those ‘insufficient’ ideas had something in them. This kind of fluency and malleability of ideas unlocks a certain invigorated innovation practice.
*I can’t locate it at the moment, but I recall reading a paper that teams had to spend something like three hours together before their ideas transcended mere repackaging of the present and really got good
Prototype Validation and Production
Vibe coding tools such as Lovable and so on, provide ways to communicate those new ideas that can be more persuasive artifacts of the experience. This goes beyond the ‘clickable’ into the ‘doable.’
Hours upon hours have been spent trying t0 make a Figma or other type of prototype appear as if it were a coded, live app or site. Copious states for an individual button or form field, so that the button could appear to respond to input from the user and changes in data from the system.
Now a vibe-coded prototype could easily project the experience with live data, created and iterated upon in in near-instant intervals.
And with this in place, the transition to Production-level code purportedly gets smoother. An AI Copilot can anticipate the ways that prototype code needs to be hardened for the product at scale.
Don’t hate, Operate
Tying this new process together is Design/Product/Dev operations. Still yet to be determined, perfected, and optimized, a new form of operations is taking shape. For example…
After an event (meeting, research call, scrum standup) an agent can transcribe, prioritize, and communicate the outputs of that call, while simultaneously creating Jira tickets for the parties that need to take action.
This is some of the automation we have started enabling at Nooma, with an agentic horizon not far behind. I’m looking forward to the oncoming changes as we all adapt to new tech.