How I AIShe built a Claude shopping assistant to stop buying cheap junk
CHAPTERS
- 0:00 – 2:33
Why shopping and returns create constant “admin work” for parents
Nicole and Claire frame modern parenting as a steady stream of online micro-tasks: purchases, support emails, and returns. They argue AI can offload this digital bureaucracy so families can spend more time on the human parts of life.
- •Parents become the “human link” between hard-to-navigate online systems
- •Shopping decisions involve hidden mental checklists and time pressure
- •AI assistance is positioned as reducing toil, not outsourcing parenting
- •Automating admin work frees time and attention for family
- 2:33 – 4:50
The core problem: panic-buying leads to cheap junk, knockoffs, and regret
Nicole explains the common cycle: urgently buying baby/kid items, defaulting to Amazon convenience, then realizing the item is low-quality, poorly made, or even a knockoff. She wants fewer, better items that can be repaired, returned, and used long-term.
- •Parents buy under stress and later regret the “crappy plastic” version
- •Concerns about third-party sellers and knockoffs on marketplaces
- •Preference for natural materials, repairability, and longevity
- •Search results are noisy: ads, SEO spam, and single-product pitches
- 4:50 – 7:43
Building a Claude Project to store household purchasing rules and trusted vendors
Nicole describes creating a dedicated Claude Project to consolidate her trusted vendor list and encode her purchasing criteria. The Project keeps shopping memory/instructions separate from other prompts and produces consistently formatted recommendations.
- •Consolidates an Apple Notes vendor list into a reusable AI system
- •Project-specific instructions prevent overfitting across unrelated chats
- •Selection criteria: heritage, craftsmanship, vendor vetting, durability
- •Avoids trendy DTC brands that may trade quality for advertising
- •Output format requirements: photo, price, materials, care notes, link, brand history
- 7:43 – 11:27
Making the “invisible checklist” visible: Claire’s anti-to-do list for mental load
Claire highlights the broader pattern: repeated purchases trigger an invisible checklist (materials, delivery timing, returns, resale, sizing). Encoding that checklist into a Claude Project reduces decision fatigue and household cognitive overhead.
- •Turn repeated judgment calls into a reusable checklist
- •Reduces mental overhead for frequent tasks happening weekly/daily
- •Works beyond shopping—any multi-step household process can be templated
- •Examples: delivery windows, returns, resale options, sizing quirks
- 11:27 – 14:29
System in action: finding a durable can opener with web search + vendor trust layers
Nicole demos a can-opener search using her Project, showing how Claude quickly surfaces a high-quality option from vetted stores. The workflow emphasizes speed (minimal prompting), comparable pricing, and fast iteration by asking follow-up questions about reviews and downsides.
- •Claude searches across preferred vendors (e.g., Boston General Store)
- •Highlights heritage brands and reliability signals
- •Shows price parity: quality items sometimes cost “Target prices”
- •Iterate by prompting: ask for downsides, review patterns, alternatives
- •Ends with ready-to-buy links from multiple reputable retailers
- 14:29 – 15:53
Can AI buy for you yet? Iteration, preferences, and the path to full delegation
They discuss whether Nicole would let agents complete purchases autonomously. Nicole isn’t fully there yet due to irregular needs and sizing complexity, but she’s moving toward standardizing recurring orders and using AI to interpret size guides.
- •Autonomous purchasing requires stable preferences and repeatable routines
- •Sizing remains a major blocker—especially for kids’ clothing
- •Claude helps parse inconsistent size guides across brands
- •Feedback loop: telling the Project what you bought improves future results
- 15:53 – 18:51
A hidden advantage: helping century-old brands with terrible websites compete
Claire and Nicole argue AI can level the playing field for small shops and legacy manufacturers whose UX and discoverability are poor. By bypassing bad navigation and search, AI makes it easier to buy from high-quality producers instead of defaulting to Amazon.
- •AI improves access to artisans and small/legacy brands
- •Old manufacturers often have the worst websites and highest purchase friction
- •“No UX is the best UX”: skip browsing and go straight to the right product
- •Reduces Amazon’s advantage that comes from convenience infrastructure
- 18:51 – 23:22
Handling failures and returns with Claude Cowork + Gmail context
Nicole switches to Claude Cowork to automate the return/refund process: find receipts in Gmail, extract order details, and draft a compelling refund email. The goal is to reduce the activation energy of these recurring 5–10 minute admin tasks.
- •Use Cowork when you need email/account access and retrieval
- •Workflow: photo of item → find receipt/order # → draft email requesting refund
- •Whisper/voice input helps when hands are busy (e.g., nursing)
- •Email includes SKU, order date, item number to avoid back-and-forth
- •Returns become fast enough that people are less likely to throw things away
- 23:22 – 25:12
Why the refund email works: surfacing manufacturing issues and review signals
Nicole explains that well-prepared refund requests often succeed because brands may already know a batch has issues. Claude can also cross-check product pages and reviews to support the claim and strengthen the request.
- •Sometimes poor durability is a known manufacturing defect
- •Claude can reference broader review patterns about quality decline
- •Stronger emails reduce additional info requests from customer service
- •Outcome: refunds (or make-goods) with minimal time spent
- 25:12 – 26:13
Using gift cards strategically to buy “classic” items that match your values
Nicole shows another query type: optimizing a purchase around a gift card or fixed budget while still meeting quality criteria. Claude tends to surface the brand’s most enduring, heritage products and provides useful provenance details.
- •Prompt: “I have $X for Brand Y—what should I buy?”
- •Add constraints: maintenance ease, price bands, longevity
- •Claude surfaces iconic, long-running products and explains why they endure
- •Useful craftsmanship/provenance snippets help decision-making
- 26:13 – 29:40
Vetting brands and spotting marketing-driven or low-quality signals
Nicole uses Claude to assess unfamiliar brands from ads by checking age, ownership changes, scaling pressures, review patterns, and operational signals. She treats it like investment-style diligence to avoid dropshipping, AI-generated reviews, and influencer-driven hype.
- •Prompt: “Is this brand legitimate and aligned with my criteria?”
- •Signals: brand age, acquisition/investment, quality changes over time
- •Look for dropshipping indicators and suspicious review language (AI reviews)
- •Marketing-heavy tactics (influencer placements) can be a yellow flag
- •Manufacturing footprint and consistency can matter for longevity/repair
- 29:40 – 33:20
Recap: the end-to-end ‘buy better’ loop—recommendations, budgets, vetting, returns
Claire summarizes Nicole’s full system: a vendor+values Project for recommendations, plus Cowork for receipts and returns. The result is fewer low-quality purchases entering the home, lower maintenance burden, and faster accountability when products fail.
- •Claude Project: trusted vendors + values-driven purchase formatting
- •Use cases: broad shopping, gift card/budget optimization, brand vetting
- •Claude Cowork: receipt retrieval + refund/return email drafting
- •Net effect: less junk, less waste, less household friction
- 33:20 – 35:35
Lightning round: how AI changes parenting—and how to correct Claude when it’s wrong
Nicole argues AI is best used to remove email-job admin tasks, not replace meaningful parenting interactions. Her prompting technique is direct: ask what went wrong, get feedback, and restate criteria like managing a coworker.
- •AI supports the human parts of parenting by removing bureaucracy
- •Automate repetitive digital tasks (returns, help emails, ordering)
- •Prompting tactic: explicitly say it misunderstood and ask why
- •Refine with new guidelines and clearer criteria rather than getting stuck
- 35:35 – 36:56
Where to find Nicole: X and Substack on technology’s role in the household
Nicole shares where she writes about ambitious household/community life and the evolving debate on AI at home. Claire closes with standard show wrap-up and ways to support the podcast.
- •Nicole on X: @nwilliams030
- •Substack interview series on tech reshaping household life
- •Discussion of how AI should/shouldn’t be integrated at home
- •Episode outro: subscribe, comment, ratings/reviews, and site info
