AI Wingmanis CustomFit's autonomous conversion-optimization layer — an AI growth team that works while you sleep. Instead of another dashboard you have to remember to check, it's a fleet of five specialist agents that continuously watch your store, your competitors, your analytics, and fresh UX research, then turn every signal into a single, evidence-backed recommendation you can ship as a real A/B test in one click.
The five agents each have a narrow job and their own cadence. AuditAgent scans your home, PDPs, and collections mobile-first and synthesizes ready-to-run experiments. CompetitorAgent diffs competitor storefronts every six hours to catch new patterns the moment they ship. BehaviorAgent runs continuous anomaly detection on your GA4 and Microsoft Clarity data, flagging conversion drops and likely root causes. KBPushAgent surfaces new teardowns and UX research matched to your vertical, and StorefrontAgent monitors Shopify every half hour so a new collection never launches un-optimized.
Everything lands as a self-contained card in one feed — no dashboards — each asking for exactly one decision: Deploy, Dismiss, or Save. Cards carry the source agent, a priority, a predicted lift, a confidence score, and evidence chips that link back to the proof. When you choose Deploy, Wingman's Copilot forks your theme, writes the variant diff, scaffolds a split-URL experiment, and wires the analytics. And because every shipped experiment becomes a labeled cohort outcome, the recommendations compound — sharper with every test, across a network of around 400 outcomes growing by a dozen a week, with an average +8.4% lift on recent ships.
What are the five agents?
AuditAgent (mobile-first store audits), CompetitorAgent (6-hourly competitor diffs), BehaviorAgent (continuous GA4 + Clarity anomaly detection), KBPushAgent (vertical-matched UX research), and StorefrontAgent (Shopify monitoring every ~30 minutes).
What does AI Wingman need to get started?
Three connections are the minimum: your storefront URL (vertical auto-detected), read-only GA4 OAuth, and at least one competitor URL. Recommended additions are the Shopify Storefront API and Microsoft Clarity. A curated knowledge base — Drip-Agency, GoodUI, Baymard, plus CustomFit's library — is built in.
How does one-click deploy work?
Choose Deploy and Wingman's Copilot forks your theme, writes the variant diff, scaffolds a split-URL experiment, and wires GA4 + Clarity events. You pick audience, split, duration, and success metric; Wingman estimates time-to-significance and notifies you when results cross the bar.
How is AI Wingman evidence-based?
Every card carries the source agent, priority, predicted lift, confidence score, and evidence chips that link to the proof. Audits stream each phase live — discover pages, scrape mobile-first, match cohort outcomes, query the KB, cross-reference GA4 + Clarity, synthesize experiments. It's evidence-first, never guesses-first.
Does AI Wingman improve over time?
Yes. Every shipped experiment becomes a labeled cohort outcome, so recommendations sharpen as the network grows — around 400 outcomes today, adding roughly 12 a week, with an average +8.4% lift on recent ships.
