A simplified view of the design process
from the Double Diamond framework
1. Discover
Understand the problem
Stakeholder interviews, user research, competitor analysis.
2. Define
Frame the challenge
Synthesize findings, define user needs, prioritise problems.
3. Develop
Explore solutions
Ideation workshops, sketching, prototyping, validation.
4. Deliver
Launch and iterate
Final design, implementation support, testing, optimisation.
Real design is almost always non-linear
While the 'Double Diamond' design framework offers a clear, four-stage structure — Discover, Define, Develop, and Deliver — it's a simplified model of what is often a much more complex and non-linear process in practice.
Design work is rarely a straight path. In reality, there's often:
Revisiting the discovery stage if assumptions prove incorrect during user validation and testing
Looping back to earlier stages when new insights surface
Redefining the problem due to shifting business goals or user feedback
Balancing constraints between user needs, technology limitations, and stakeholder priorities
Iterative refinement based on usability testing, stakeholder input, or implementation challenges
The Double Diamond is a helpful framework, but real design work is messy, iterative, collaborative, and requires continual adaptation.
AI in the Design Process
AI does not remove the need for good design process. It changes where time is spent, what gets accelerated, and what new risks appear. Using the Double Diamond as a guide, AI tends to speed up execution more than judgment. The earlier stages still depend heavily on human framing, critical thinking, and stakeholder alignment, while later stages benefit more from rapid generation and iteration.
1. Discover
Still essential, but faster to process and broaden
This stage remains critical because AI cannot reliably tell you which problem is worth solving unless the inputs, context, and organisational constraints are already understood.
AI is useful here for:
- summarising interviews and workshops
- clustering themes from research notes
- extracting patterns from survey responses
- generating draft research plans, scripts, and hypotheses
- speeding up desk research and competitor scans
But there is a catch: AI can make weak research look polished. It can summarise what people said, but it does not automatically understand what matters most, what is contradictory, what is missing, or what is politically sensitive in an organisation.
So Discover becomes:
- less manual
- more synthesised
- potentially broader in coverage
- but still dependent on human interpretation
The risk is that teams may confuse faster synthesis with real understanding.
2. Define
Arguably even more important in an AI-enabled process
Define is where the value of the designer may increase, not decrease.
If AI makes it easier to generate solutions quickly, then it becomes even more important to:
- frame the right problem
- define success clearly
- prioritise user needs vs business needs
- identify constraints, risks, and trade-offs
- write sharp problem statements and design principles
Without strong Define work, AI just helps teams move quickly in the wrong direction.
Discover and Define do not disappear. In many cases they become the quality control layer for everything that follows. AI can help compress the mechanics of these stages, but not replace the strategic thinking.
3. Develop
This is where compression becomes most visible
This is where AI has the strongest immediate impact.
AI can help teams:
- generate many ideas quickly
- create user flows, wireframes, content drafts, and interface variants
- turn prompts into code prototypes
- simulate edge cases or alternative journeys
- produce components and interaction ideas faster
- test multiple concepts before investing heavily
This changes the nature of design work from manually producing each artifact to directing, editing, evaluating, and combining outputs
So the designer's role shifts more toward:
- setting intent
- curating quality
- spotting weak assumptions
- maintaining coherence across journeys and systems
The benefit is obvious: more concepts, faster iteration, lower cost of exploration.
The risk is also obvious: teams may generate lots of plausible-looking designs that lacks depth, consistency, accessibility, or real user value.
4. Deliver
Also compressed, but with new review burdens
Delivery can also accelerate because AI can support:
- faster prototyping
- front-end code generation
- component documentation
- microcopy and content production
- design-to-code handoff
- test case generation
- accessibility checks
- design system usage guidance
But Deliver is not just “make it quicker.” It still needs:
- design QA
- technical feasibility checks
- governance
- content accuracy
- brand consistency
- measurement after release
In other words, AI may reduce production effort, but it often increases the need for review and verification.
My overall view
AI compresses production more than it compresses thinking.
Or even more specifically:
- Discover: accelerated by AI, but still human-led
- Define: still essential, perhaps more essential than before
- Develop: heavily accelerated and expanded by AI
- Deliver: accelerated, but requires stronger validation and QA
Discover and Define remain essential for identifying the right problems and framing them well, while AI shortens research administration and synthesis. The biggest acceleration happens in Develop and Deliver, where AI helps teams generate, prototype, and refine solutions much faster. As a result, the design process becomes more compressed and iterative, but human judgment remains critical throughout.