Andy Ke

AI and growth product manager, Taelor and Kesha Select

I ship AI that experts use, and run growth with self-improving agents.

At Taelor I build AI tools for professional stylists, with measures and fallbacks built in. At Kesha Select I run growth through an AI agent system I built: it runs daily checks and growth reviews, applies routine fixes itself and brings major changes to me.

45%
faster styling

Taelor. The AI pairing workflow I launched, used by 90% of stylists.

+60%
recommendation success rate

Taelor. AI quality guardrails in 2026: stylists approved more of the AI's recommendations, with zero customer-reported sizing errors.

+41%
daily orders

Kesha Select. Within three months, store-wide while I ran growth, with monthly revenue up 57%.

BeforeGain
Senior PM, TaelorProduct Manager, TaelorGrowth consultant, Kesha Select

At Kesha Select I report to the CEO and run growth strategy and execution with the store's cross-functional team. About 10 years across engineering and product.

Taelor

Taelor is a men’s clothing rental startup, featured in GQ and Vogue and at TechCrunch Disrupt.

AI for stylists

Styling 45% faster with the AI workflow I launched

I defined the requirements, the measures and the AI-assisted workflow for Taelor's new styling tool. Quality thresholds and fallback rules decide what reaches a customer; the stylist makes the final call.

45%
faster styling
90%
of stylists adopted it
+60%
stylist-approved recommendations, zero customer-reported sizing errors

How I built it. I used ChatGPT for prototyping, ideation and analysis; I owned the requirements, pairing rules, guardrails and measures. Engineering built the production tool.

Read the AI case
Redrawn PairsWith screen: a picked item, ranked AI matches and stylist ratings
PairsWith, the AI pairing workflow, redrawn from the shipped tool. Names and details are fictional.
  1. The stylist picks an item.
  2. The AI suggests matches, ranked by rules built from outfits customers kept.
  3. Stylist ratings shaped each rules release.

Homepage rebuild, interviews to release

I led a cross-team rebuild of Taelor's busiest page, and every section shipped

Taelor's homepage got the most traffic and converted below benchmark. I ran many customer interviews, ranked the four doubts first-time visitors raised, and led designers, copywriters and a 10+ person team to rebuild the page around them. Every section shipped and was measured section by section.

Interviews
many customer interviews behind the brief
Cross-team
design, copy, data, merchandising, engineering and ops
Per-section
click-through and drop-off read for each section
Read the homepage case
The rebuilt Taelor homepage, redrawn, with numbered markers tying each section group to a visitor doubt
The rebuilt homepage, redrawn. Each numbered marker ties a group of sections to the visitor doubt it answers.
  1. Is this a rental, and how is it different? The hero and Why Subscribe To Taelor.
  2. Do you have styles I like? The occasion tiles and Stylish Looks by Our Expert Stylists.
  3. Can I trust you? Style in Action: Real Customer Experience and Trusted by Our Customers.
  4. Do you have my size and brands? The Hundreds of Brands & Styles headline and a link to the size page.

AI Style Report, live

A free AI style report you can try right now

Tap the looks you'd wear and the AI returns your style direction, colors and outfits. The full report unlocks with an email. I built it end to end.

Live
on Taelor's public site
95%
of starters finish the quiz in an early read, vs 55–69% in industry benchmarks

How I built it. Built with Claude and Cursor: I designed and built it end to end.

Step 1: pick where you need outfitsStep 2: tap the looks closest to your styleStep 3: your style and first outfit appear
Captured from the public live site, report.taelor.style.
  1. Pick where you need outfits. No photo, no card.
  2. Tap the looks closest to your style; the AI builds outfits from stylist-approved looks.
  3. See your style right away; the full report unlocks with an email.

Kesha Select

AI agent operations

An AI agent system that runs daily operations and audits itself

Agents check and draft across the store, search, email and ads every day. Each day the system reviews its own work, applies routine fixes itself and brings major changes to me; only one agent can write.

1Daily jobsNightly site checks; scheduled search, email and ads reviews
Specialist agentsRead only. Store, search, AI answers, accessibility, email, ads
2I approveSpend, discounts, emails and customer-facing changes
One writerThe only agent that can change the store; every write backed up
LiveStore, ads, email
3Self-improving loop. A daily audit reviews every job and proposes changes to the system itself.
How a change reaches the store: who starts the work, who can write, where approval sits.
  1. Work starts on its own: nightly site checks, plus scheduled search, email and ads reviews.
  2. Drafts arrive as proposals; spend, discounts, emails and customer-facing changes wait for approval.
  3. The dashed loop: the system correcting itself; major changes come to me.
Daily self-audit
Single writer every other agent only reads
My approval on major changes

How I built it. Built on Claude Code: I designed the system, agents and approval rules; agents apply routine corrections themselves, and major changes wait for me.

See how the system works

Growth consultant to the CEO, strategy to release

Daily orders up 41% and monthly revenue up 57% in three months

I led the store's cross-functional team, reporting to the CEO. Session recordings, customer interviews and funnel data showed where carts stalled: hidden campaign minimums. I replaced them with a single, clearly explained free-shipping threshold and ranked every other fix by cost. The order and revenue gains are store-wide, not isolated to the cart changes.

+41%
daily orders, store-wide
+57%
monthly revenue
+30%
cart-to-checkout conversion, associated with the cart changes

How I built it. Built with Claude Code: an AI-assisted workflow helped me ship 10+ prototypes in a month; I owned the research, specs, release plan, QA and dashboards.

Read the checkout case
Cart drawer before and after: three campaign minimums become one free-shipping bar
The cart drawer before and after, recreated in English. Prices are illustrative.
  1. Hidden per-campaign minimums became a single threshold with a bar that names the gap.
  2. Recommendations left the cart; shoppers tapped them and never came back.
  3. Split orders merge, so shoppers stop paying shipping on each one.

My own product

My own product, v0.2 preview

Pocket Tennis Team: a personal tennis coaching team on Claude

A Head Coach you talk to, backed by six specialists, each a Claude skill. For players and parents; it runs on the player's computer today, with a live demo. CourtPlan is for coaches and still in discovery.

About 3 min to about 20 s
coach reply time, in my own end-to-end tests with the live AI
7
team members: a Head Coach plus six specialists, each a Claude skill

How I built it. I designed and built it end to end with Claude; the source is public on GitHub.

Pocket Tennis Team demo: the interview with the plan taking shape beside it
The live demo, with sample data.
  1. Interview with a live plan board: As the player answers, the plan builds on screen right away, without waiting on the AI.

My own product, pre-pilot

CourtPlan: an assistant coach that remembers every student

Coaches who run small groups plan at the fence and keep progress in their heads. I sized the market, mapped 12 tools, designed the system and built a clickable prototype. Next is a hand-run test with 2 pilot coaches.

12
tools mapped, one open gap
2
pilot coaches lined up for a hand-run test

How I built it. I did the market research, the competitive scan, the design brief and the clickable prototype solo with Claude Code.

Read the CourtPlan case
Prototype: check-in, who's here tonightPrototype: tonight's plan per courtPrototype: live session timer
From my clickable prototype, with example data.
  1. Who's here tonight?: The coach taps who showed up and sets the courts. No fees appear before or during class.
  2. Tonight's plan: Groups by level on each court, with drills from the coach's own library and one cue per block.
  3. Live session timer: A countdown for each block, the drill and cue on every court, and quick notes per player.

Hiring an AI or Growth PM?

I'm looking for full-time AI PM, Growth PM and technical PM roles. Email is fastest.

Run a Shopify store? Ask about a growth audit.