Where should AI live while you're designing a product, and after you've shipped it?
Threadwise tells a shopper if a piece of clothing is worth buying, based on its fabric. That's the backdrop for two decisions.
First: to design the prototype, three AI models proposed different directions, and I directed AI to combine the strongest parts of each into the final design.
Second: the product became research into cutting the cost of AI itself. The answer was that expensive AI reasoning should happen once, offline, never on every request while someone's using the app.
The final, human-curated Threadwise prototype. Fable's design identity, Sonnet's completeness, and Opus's discipline, rebuilt into one flow. Tap through it above.
ROLE
METHOD
TOOLS
HEADLINE OUTCOME
Product Designer, solo
AI-architecture design + 3-model controlled exploration
Zero runtime AI calls, the intelligence is frozen at build time
Claude Design — Sonnet, Opus, Fable
Where should AI live while you're designing a product, and after you've shipped it?
01 THE REAL EXPERIMENT
Two moments, two decisions, and for Threadwise, both came down to the same rule: AI proposes, I decide.
While designing. Three Claude models (Sonnet, Opus, and Fable) each proposed a full direction for the product. I compared what they made, then directed AI to combine the strongest parts into one final design.
After shipping. The product itself never calls AI. Fabric data gets researched and scored once, checked by a person, then frozen into the app. A shopper opens Threadwise and gets an instant answer, with nothing running live.
A product that calls an AI model on every user request is slower, more expensive, and more fragile than it needs to be. And every answer is a fresh roll of the dice. Freezing the expensive reasoning at build time, and verifying it once by hand, makes the runtime experience instant, private, consistent, and cheap to run at any scale. Every verdict traces to a reviewed table, so the product is auditable in a way per-request inference never is.
There's a quieter point underneath: a sustainability product shouldn't burn compute re-deriving the same answer millions of times. Threadwise applies its own ethic to its AI: use exactly as much as the job needs, exactly once.
why this matters
A shopper, a label, and a decision with seconds to spare.
02 The context
Threadwise helps a shopper decide, garment in hand, whether it's worth buying, based on what its fabric composition actually means for comfort, durability, care, microplastics, and end of life, not just the percentages printed on the label.
Alongside that product problem, I used Threadwise as a second, deliberate experiment: could AI support product design across strategy, exploration, prototyping, and evaluation, without ever being the one to make the final call? This case study is both things, in that order.
The label was honest. It just wasn't useful.
Garment labels give shoppers percentages, not judgment. Knowing an item is 80% cotton and 20% polyester doesn't tell you how that blend will actually behave: how it wears, how it washes, whether it sheds microplastics, or what happens to it at the end of its life. Sustainability information exists, but it's scattered, jargon-heavy, and never present at the one moment it would change a decision.
03 The problem
How might Threadwise translate fabric composition into a quick, honest, non-judgmental purchase decision?
Seven documents, before a single interface.
04 FOUNDATION
I directed these seven documents into existence rather than writing them solo. I set the thinking and made the judgment calls, then handed execution to whichever model suited the task: Sonnet for the core product goal and problem framing, Opus for structural and visual direction, Fable for the scoring logic's fiber math. Opus then compiled everything into the documentation site linked below and froze it there.
This is where the product's real judgment calls got made: what the score can and can't see, what stays out of the MVP, what tone the product refuses to take, so no model would have to invent them later. The freeze had an economic job too: a settled brief means exploration happens once, against stable ground, instead of being regenerated every time the thinking moves.
Where AI made this faster and where it didn't get a vote
Building the product surface: AI helped rapidly build the website itself, the expectation-setting layer (who the practice is for, how classes run, what each session includes) that turned manual explaining into self-serve onboarding.
Content production: first drafts of class content (Yoga Nidra scripts, sequences, opening and closing scripts), class communications, social content, and service copy, all rewritten to my voice and standard.
Competitive scans: first-pass landscape mapping across 15+ yoga and wellness services before I went deep manually.
Synthesis: clustering inquiry threads, student feedback, and competitor notes into candidate themes, hours instead of days.
Repeatability: reusable prompt workflows for recurring analysis, so rigor didn't depend on my energy that week.
05 The AI-augmented workflow
AI ACCELERATEd
I DECIDED
This project is where my AI-augmented process became a documented system rather than an experiment. The honest accounting:
The positioning call: small-batch premium over volume, a strategy decision AI can argue both sides of, which is exactly why it can't make it.
The framing: that this was a service-system problem, not a marketing problem.
The voice: every student-facing word. Warm, brief, and mine.
Verification: every AI-assisted output checked against the source data before it informed a decision.
Measured on a living practice
06 Outcomes
50–60%
2 months
5–6 mo.
reduction in onboarding friction. Structured flows replaced open-ended message threads; inquiries arrive pre-informed.
from launch to converting free pilot participants into paying customers, validating product-market fit on real money.
average student retention through iterative, feedback-driven service improvements.
From zero
~3 hr/wk → 0
faster research synthesis and content production via the documented AI-augmented workflow.
SEO discoverability and a three-platform content system (Instagram, Facebook, YouTube Shorts) built from nothing.
the weekly manual explanation loop eliminated; the website now carries the pre-enrollment conversation.
Everything described above is currently running on vaidehiyoga.com.
~40%
07 What this project proves
Why a living practice belongs in a product portfolio
Because nothing here was hypothetical. Every persona was a person who paid me or didn't. Every flow was tested by someone who could simply leave.
What I'd do next: Phase 2 automates the backstage, enrollment confirmations, payment reminders, and scheduling, while keeping every human touchpoint human.
Next case study →
© 2026 Vaidehi Yelkawar