By 2025 customer research was moving into AI assistants, and Cox’s most important pages were invisible to the crawlers assembling the answers. I built the GEO and AEO practice from zero: an AI search landscape POV, a noscript content workflow, an llms.txt implementation guide, and a 37-item dual-crawl QA protocol. The result was a 10 to 15% citation lift and Cox positioned as the most-cited telecom in unbranded AI search.

Cox’s chatbot and call transcripts held the richest dataset about customer friction, but nobody could use it. I led the product from pre-POC through Call Pipeline V2: authoring requirements, introducing affinity mapping as the taxonomy method, and designing the human-in-the-loop validation that took unknown classifications from 28% to under 2% across a 79-code, three-dimension taxonomy.

Over 60% of Discover customers wanted budgeting and cash flow tools but lacked personalized guidance. As lead experience designer and strategist I shaped a financial wellness suite built on Personetics: a customizable dashboard, an Insights Inbox, budgeting, and spending trends, with a redesigned landing experience that significantly improved understanding and opt-in rates.

A free Discover service that scans ten people-search sites and removes personal data over a 91-day quarterly cycle. The design challenge was legibility: users would assume the scan ran on every visit. I led UX from a two-screen onboarding through a status-driven dashboard with drill-down scan details and distinct states for in-progress removals, completed removals, and a finished quarter.

AI assistants affirm users about 49% more often than humans do, and users like it, which is exactly the problem. An essay on sycophancy as a measurable training artifact, why memory and personalization compound it, and why the cheapest fix, testing before shipping, is the one most teams skip.

Only about half of AI citations fully support their claims, the web is not equally crawlable across engines, and nobody clicks through anymore. An essay on citation faithfulness, differential crawlability, collapsed crawl-to-referral economics, and who controls what you read. Written with 24 sourced footnotes and five custom charts.

A homebrew D&D world run with AI-assisted lore, Midjourney concept art, and custom Foundry VTT macros, now becoming a Golden Axe-style brawler built in Godot 4 with Claude Code: a 2.5D floor with Y-sorting, hitbox and hurtbox combat, a magic aura buff, and an animated title screen. Exploratory AI practice pointed at something joyful.