DesignTech
EXP#
10
Solvr
A guided design workspace that takes a product problem through seven structured stages — Project Setup, Discover, Define, Ideate, Solution, Validate, Iterate — with AI drafting, critiquing and scoring readiness at each one.
02
Context
Most AI design tools compress a product problem straight into a finished-looking output — a generated spec, a set of screens — with no visible working. Solvr set out to do the opposite: guide a problem through the same stages a structured design process would use, keeping evidence, assumptions and gaps visible at every step instead of hiding them behind a single generate button.
03
The Challenge
The product needed to feel like a design partner that pushes back, not a generator that produces a confident-sounding answer to whatever's typed in — while still being usable by someone who doesn't already know a formal UX methodology, and without ever letting the process feel like a rigid form to fill in.
04
Research
Reviewed how existing AI design tools collapse a problem straight into an output, and found almost none separated evidence from assumption — which became the central design bet for Solvr instead of a secondary feature.
05
Insights
The key findings that shaped the direction.
01
A generate button isn't a design partner
Tools that produce a finished-looking spec from one prompt skip the part where a real design process would surface what's still unknown — Solvr needed to slow that down, not speed it up further.
02
Evidence and assumption have to stay visibly separate
The moment an assumption gets treated like a fact, every decision built on top of it inherits the risk — so every stage explicitly tags what's evidence, what's assumption, and what's still a gap.
03
Readiness has to be scored per stage, not once at the end
A single overall completeness score hides exactly where a project is weak — scoring each of the seven stages independently is what makes 'are we ready to move on' an answerable question.
06
How decisions were made
01
Structure the whole process around seven named stages
Built Project Setup, Discover, Define, Ideate, Solution, Validate and Iterate as distinct, sequential stages, each with its own inputs and its own definition of done.
02
Separate evidence from assumption at every stage
Gave every stage explicit Strengths, Gaps and Critical Assumptions panels, so a project's confidence level is always visible next to its content, not implied by how polished the output looks.
03
Score readiness per stage, not the project as a whole
Attached a readiness percentage and a plain-language 'what's needed next' note to each stage individually, so moving forward is a decision backed by a specific gap list, not a vibe.
04
Make the AI challenge the framing, not just answer it
Directed the AI to flag gaps and question assumptions at every stage rather than accept the first framing of a problem, so the tool behaves like a partner pushing back, not an assistant agreeing along.
07
Solution
A guided design workspace that takes a product problem through seven structured stages — Project Setup, Discover, Define, Ideate, Solution, Validate, Iterate — with AI drafting, critiquing and scoring readiness at each one.
Per-stage readiness scoring with named Strengths, Gaps and Critical Assumptions
Seven-stage guided process from problem framing through to an evidence-backed iteration plan
Structured, exportable product and UX specification generated from the Solution stage
AI guidance panel that recommends what to resolve before advancing to the next stage
.png)