Know what changed, what it costs — and what to fix first
Copilot scans your storefront the way a shopper would, reads the store’s connected Shopify, behaviour and module data, and turns it into one CRO score, a written daily brief and a ranked queue of evidence-backed actions — each with confidence, risk, a proposed change and a measurement plan. Apply, measure, revert. Missing inputs are named, never assumed.
- Read-only scan — never changes your store
- Missing inputs shown explicitly, never assumed
- Pause or revert any change · complete audit trail
Mobile add-to-cart trails desktop by 4.5 points and the cart drawer hides the free-delivery threshold. Two low-risk, reversible changes cover most of this month’s gap.
Based on storefront scan · 12 store signals · Manual approval
60–90 seconds · nothing was applied to your store
Autopilot guardrails
60–90 s
Read-only storefront scan
Homepage, catalogue, a sample product page and the cart — read the way a shopper would, nothing applied
70+
CRO score target
A 0–100 score with a six-part weighted breakdown; “Healthy” starts at 70
17
Data sources, each shown live or not
Every source is marked Connected or Unavailable — Copilot only reasons over what is connected
3 / day
Autopilot cap by default
10 per week, minimum 80% confidence, low risk only, reversible only — checkout always blocked
A read-only scan of your storefront, in 60–90 seconds
Click Run AI scan and Copilot opens your homepage, a product page and the cart the way a shopper does, then scores what it finds against the same CRO evidence the rest of the dashboard uses. Four passes — connect, read, score, quantify — then Copilot writes the brief. Nothing is applied to your store; the scan reads public storefront pages only.
- A CRO score from 0 to 100 with a six-part weighted breakdown: homepage clarity, PDP structure, cart and checkout friction, trust signals, offer and AOV system, technical basics
- Top conversion issues with recommended fixes, detected opportunities, missing signals and a directional monthly revenue-leak range — labelled as an estimate
- The score trends across your last six scans on the command center, against a dashed “Target 70” line
What this scan does
Four storefront passes, then your brief
Copilot opens your homepage, a product page and the cart the way a shopper does, then scores what it finds against the same CRO evidence the rest of the dashboard uses.
This usually takes 60–90 seconds. Stay on this page.
Every morning: what changed, what it costs, what to fix first
At 04:30 UTC every day — and on demand after any scan — Copilot reads the full store context: metrics against the previous day and week, top products, orders, behaviour, traffic, module status, the CRO snapshot and every recent recommendation. It returns a written brief of at most three or four insights, opportunities and alerts, plus up to five fresh recommendations, each with category, priority, confidence, evidence, proposed change, risk and whether Autopilot may touch it.
- Key metrics with “vs prior day” or “No comparison” — Conversion Rate, Average Order Value, Revenue per Visitor, Cart Abandonment, Sessions, Add to Cart Rate and more; unavailable metrics print “Unavailable”
- “Do this first”: the highest-impact opportunity with Priority 01, confidence and monthly potential, then 02, 03, 04 in sequence
- “Signals to verify” — missing inputs are shown explicitly. Copilot will not replace them with assumptions
Comver AI Commerce OS
AI Commerce Copilot
Daily brief · Aug 16, 2026, 06:30
Mobile add-to-cart trails desktop by 4.5 points and the cart drawer hides the free-delivery threshold. Two low-risk, reversible changes cover most of this month’s gap.
Based on storefront scan · 12 store signals · Manual approval
Conversion Rate
2.6%
−0.3 pts vs prior day
Average Order Value
$74
+2.1% vs prior day
Cart Abandonment
68%
+1.4 pts vs prior day
Mobile Sessions
61%
No comparison
Priority 01
88%
confidence
Do this first · Highest-impact opportunity
Add a sticky add-to-cart bar on mobile product pages
Mobile add-to-cart rate 8.9% vs desktop 13.4% (Comver storefront events). Sticky ATC is an inactive Comver module — activate it from Revenue modules.
Connected data sources
14 of 17 connectedHeatmaps · Unavailable — no behaviour events in the comparison window. Copilot will not replace them with assumptions.
Ranked, evidence-backed. Apply, measure, revert.
One queue holds open recommendations and in-flight changes, sorted priority → impact. Prepare change turns a recommendation into an exact, reviewable proposal; every applied change is measured against its own plan and can be reverted in one click.
Every recommendation with its stage, confidence and estimated impact
Filter by priority — critical, high, medium, low — and by status. Each row shows the category, the page it touches, confidence and monthly impact (or “Not estimated”), and its stage: Detected, Suggested, Ready to review, Manual required, Approved, Scheduled, Live, Measuring, Completed, Failed. Anything not controlled by a Comver module becomes “Manual required” — Copilot never edits your theme behind your back.
Comver AI Commerce OS
AI Commerce Copilot
Sorted priority → impact · Stages: Detected · Suggested · Ready to review · Manual required · Approved · Scheduled · Live · Measuring · Completed · Failed. Anything not controlled by a Comver module becomes “Manual required”.
Ask what changed, why it matters, and what to do next
Ask Copilot is a chat grounded in your connected data — current Shopify, Comver analytics, modules and audit context, with a 30-message memory per conversation. It answers with the numbers and names the source of each one; when a source is unavailable it says so instead of guessing. Six suggested questions get you started, and the floating Copilot launcher carries route-aware prompts on every page.
- “Why did conversion change this week?” · “Show me the largest revenue leaks.” · “Compare mobile and desktop performance.” · “What changes did Copilot make in the last 7 days?”
- Answers cite Shopify orders, Comver sessions and events, module config and the latest audit — and mark anything Unavailable
- The same grounded chat opens from the floating Comver Copilot orb, with prompts that match the page you are on
Comver AI Commerce OS
AI Commerce Copilot
Suggested questions
Grounded in connected data — Copilot names unavailable sources instead of filling gaps with assumptions.
Conversion fell from 2.9% to 2.6% week over week (Shopify orders + Comver sessions). Two things moved:
- Mobile sessions rose to 61% of traffic, and mobile add-to-cart is 8.9% against 13.4% on desktop (Comver storefront events).
- Cart abandonment is 68%, up 1.4 points; 31% of abandoned carts sit $8–15 below the $60 free-delivery threshold, which the drawer does not show (Cart Rewards config).
Heatmaps: Unavailable for this window — no click data was recorded, so I am not attributing the change to a layout cause. The two actions above are already in your queue (01 and 02).
Automation with a policy, not a personality
Autopilot is off until you switch it on and tick “I understand that AI may automatically apply selected changes.” Then it applies only what your policy allows: chosen categories, a risk ceiling, daily and weekly limits, a confidence floor, blocked pages and modules and a campaign freeze for sensitive dates. It can only ever execute reversible changes on Comver-controlled modules — Bundling & Upsells, Offers and Cart Rewards. Everything else waits for you.
- High-risk areas always stay manual: Checkout Flow, Speed / Technical and Pricing / Value Communication are locked — never auto-applied
- Defaults: low risk only, 3 changes a day, 10 a week, minimum 80% confidence, checkout page blocked, paused during active tests
- A complete rollback snapshot is required before anything applies — pause or revert at any time from History
Comver AI Commerce OS
AI Commerce Copilot
Autopilot active — 3 categories allowed
Automatically applies approved categories of low-risk changes. High-risk areas always stay manual. You can pause or revert changes at any time.
Auto-apply categories
14 categories · risk labels from the Copilot contract
Risk policy
Safety limits
From the first scan to a measured result
No theme edit and no separate app. Copilot lives in the Comver workspace and is reached from the command center, the floating orb and the Action Plan.
- 1
Connect
Install the Comver Shopify app. Products, orders and customers arrive from the Admin API; switch on storefront tracking and sessions, events, heatmaps and funnels join the 17 data sources.
- 2
Scan
Run AI scan from the command center or Scan store in Copilot. Sixty to ninety seconds later you have a CRO score, the top issues, a leak range marked as an estimate — and the first brief.
- 3
Decide
Prepare change on any recommendation, review the evidence and the before/after, then Reject, Edit, Schedule, Approve or Apply now. Or let Autopilot handle the low-risk, reversible ones inside your policy.
- 4
Measure
Applied changes move to Measuring. After at least 24 hours and 50 sessions on each side, Copilot compares before and after on the plan’s primary metric, records the uplift and logs it — or you revert.
Estimates say so. Gaps say so.
Copilot keeps facts, estimates and unknowns visibly apart. It never fills a gap with an assumption, never applies a change it cannot roll back, and never touches pricing, checkout, payment, shipping, legal, policy or tracking on its own.
≈
Every dollar figure is an estimate until measured
Monthly potential, revenue at risk and leak ranges are directional — and say so
0
Assumptions where a source is missing
“Missing inputs are shown explicitly. Copilot will not replace them with assumptions.”
3
Categories locked to manual, always
Checkout Flow · Speed / Technical · Pricing / Value Communication
Copilot executes only on Comver-controlled modules (Bundling & Upsells, Offers, Cart Rewards); every other target is prepared for you to carry out. Copilot has no Beta badge and no separate price.
Frequently asked questions
Nothing. It is read-only: it fetches your public homepage, product catalogue, a sample product page and the cart, takes screenshots and scores what a shopper would see. If anything fails you see “Nothing was applied to your store.” No theme app embed is needed for the scan, the brief or the CRO score.
About 60–90 seconds. You get a CRO score from 0 to 100 with a six-part weighted breakdown, the top conversion issues with recommended fixes, detected opportunities, missing signals and a directional monthly revenue-leak range — plus, from Copilot, a written daily brief and a ranked action queue.
One canonical estimator used on every screen: the larger of the latest audit’s leak range and the sum of dollar impacts on open recommendations, capped by measured monthly revenue when it is known. If neither source exists the UI says “Awaiting data” or “Run a scan to quantify” rather than inventing a figure.
Only if you turn Autopilot on, tick the consent checkbox, and only for the categories and risk level you allow — low-risk, reversible changes to Comver-controlled modules (Bundling & Upsells, Offers, Cart Rewards) above your confidence threshold, within daily and weekly limits and outside campaign freezes. Checkout Flow, Speed / Technical and Pricing / Value Communication are always manual; high and critical risk always need your approval.
Yes. Every applied module change keeps a rollback snapshot; a snapshot is required before Autopilot may act at all. Revert change restores the previous configuration, stops the measurement and logs the revert in the History timeline.
Each change ships with a measurement plan — primary metric, minimum window, success and rollback conditions. After at least 24 hours and 50 sessions on each side, Copilot compares before and after on that metric and records the uplift; the timeline logs “Measurement completed: … changed from X% to Y%.” Behaviour tracking must be live for measurement; otherwise it waits for data rather than guessing.
Shopify products, orders and customers; Comver storefront sessions, events, heatmaps, funnels and traffic sources; revenue-module status; the CRO score, revenue leaks and storefront audit; AI CRO audits, experiments, personalization and segments. The Connected data sources grid shows exactly which of the 17 are live for your store — anything missing is labelled Unavailable, never assumed.
Automatically every day at 04:30 UTC for every connected store, and on demand whenever you run Scan store. The daily run also applies scheduled changes and measures live ones. Copilot has no sidebar row of its own: open it from the command center (“Ask Copilot”), the floating Comver Copilot orb on every page, the command palette or the Action Plan.
Let Copilot read your storefront this morning
Install Comver, run the read-only scan and get your CRO score, brief and first ranked actions in about ninety seconds. Apply what you agree with, measure it, revert what you do not.
14-day free trial · Live on Shopify in minutes · Checkout stays in Shopify