Your store, run by
systems that learn.

Four verbs. We ship all four.

Most “AI for ecommerce” is content generation. Ours runs the parts of the business nobody has time for — the retries, the reconciliations, the reorder timing, the tickets. 12 systems, grouped under four verbs.

• 9%

  Recover

  Money you already earned, going out the door at the payment layer.

  9% of recurring revenue goes to cards that decline and are never properly retried. Recharge, Skio and Smartrr all ship a flat schedule that treats every failure the same. We replace it with routing that reads the decline code first — insufficient funds is a different problem from an expired card — and picks the retry window, the dunning message and the card-update path per code.

  • Payment recovery →
  • Fraud & chargebacks →

• 41%

  Retain

  Money that repeats — if the second order happens.

  41% of email revenue comes from automated flows, off about 5% of sends. Underneath ours: a model that sets the reorder prompt from actual consumption per SKU rather than a flat 30-day guess, and a cancel flow that offers a pause, a smaller quantity or a later date instead of a cancel button. Typical monthly subscription churn is 6–7%, and the gap to top quartile is almost always billing infrastructure and cancel-flow logic.

  • Lifecycle flows →
  • Subscription retention →
  • Post-purchase & AOV →

• 12+

  Run

  Money you spend on people doing what a system should do.

  12+ hours a week go on reconciling Shopify against Amazon, the 3PL against the returns queue, the ad platforms against the P&L. Agents on n8n running on your own VPS do the reading and the drafting: order status answered from the actual fulfilment record, the purchase order drafted off real velocity and supplier lead times, a nightly job that reports the rows where two systems disagree. Anything touching a refund, a discount or a cancellation is prepared by the agent and pressed by a person.

  • Ops automation →
  • Customer service AI →
  • Catalog & feeds →
  • Reporting & analytics →
  • AI agents & automation →

• 13×

  Reach

  Money that never arrives, because the model never named you.

  Orders from AI-powered search on Shopify grew about 13× year over year. Whether ChatGPT, Gemini or Perplexity names your brand when a customer asks for a recommendation is now a distribution question, and almost nobody in this revenue band can tell you their current answer. We run the prompt set on a schedule to get a baseline, then fix what those models actually read: structured data on real templates, one canonical entity across the web, and the third-party surface they quote.

  • AI search visibility →
  • Paid media →

Sources, in order: Baremetrics · Klaviyo benchmark data, 183,000+ brands · Recurly Consumer Goods Churn Benchmark, for the 6–7% churn figure inside Retain · Harley Finkelstein, Shopify Q1 2026 earnings call, for the 13×. The manual-ops figure under Run is our own estimate for a brand at this size, not a published benchmark — we label it that way rather than dress it up as one.

What this looks like in practice.

Public teardowns of real brands’ retention and ops setups — the parts you can observe from the outside: post-purchase email, cancel flow, reorder timing, and what a model says when you ask it to recommend a brand like theirs.

• Teardown 01 · Supplement subscription

  —

  Recovered failed payments in 60 days metric to confirm

  Brand and result to be published once the teardown is written. Nothing here is a placeholder for a number we don’t have.

• Teardown 02 · Cold-chain food & beverage

  —

  Reorder-window accuracy metric to confirm

  Brand and result to be published once the teardown is written.