Straight answers.

A lightweight, asynchronous script (or a Shopify app) plus read-only analytics connections. No meaningful page weight, nothing render-blocking, and nothing customer-facing changes until you approve it. The agents observe, verify tracking in a real browser, and deploy approved experiments and segment rules through your testing layer.

The agents build a model of your funnel — margins, catalog, traffic mix, audience behavior — and generate the roadmap from that model. A high-AOV subscription brand and a low-AOV impulse brand get different tests, different segments, different guardrails. No template playbooks.

Your choice of mode. Suggest: every experiment and segment rule waits for your approval. Autopilot: agents launch within pre-agreed guardrails. Either way, every test carries a pre-registered primary metric, guardrails (AOV, refund rate, ROAS) and a ship / iterate / kill rule.

Visitors are segmented in real time — source & campaign, new vs. returning, device, cart value, behavior — and each segment can get a personalized journey, not just a page: you define the path (up to 10 pages), the Segment agent adapts copy, imagery and offers per page, you approve everything in one review, and the flow activates when visitors follow that path. Every journey runs against a holdout with real-time drop-off analytics, so the lift is measured, not assumed.

First-party data only, consent-aware, no PII sold or shared — and one client's data never trains or informs another client's agents. Workspaces are fully isolated. See our privacy policy.

Full A/B velocity needs roughly 30K+ sessions/month. Below that the platform runs sequential (pre/post) testing, segmentation still works — and we'll tell you honestly which mode fits your traffic instead of selling underpowered tests.

The audit. Agents scan your funnel, verify your tracking, price every leak, and produce the roadmap — credited to your first plan month if you continue. See pricing.