Autopilot runs A/B tests on your live checkout, retries the payments that fail, prepares the disputes that land, and tells you what it did. Every rule is a switch. Nothing is sent without your say-so.
Autopilot sits in the header of each portal page. Open it and you see what Krepling does there on its own, one switch per rule, with a pause for the whole page.
2 cases are due within 3 days, $214.00 at stake. Krepling has prepared a response for each: what to send, the proof and the letter. Check them and send, or let Krepling send them.
Default rules are housekeeping you'd never want to do by hand, and each one leaves a line in the log.
Rules that act in your name are off until you turn them on, and say so in plain words.
One switch pauses every rule on a page. Settings are kept for when you flip it back.
Field order, wallet order, a shipping threshold, a trust line. Autopilot splits your real shoppers, counts who pays on each side, and keeps the winner only once it's sure.
Half see A, half see B. Same products, same prices, one difference.
A result counts when the probability clears your line. Until then both keep running.
Every kept change sits in the log with an undo, and the losing version is a click away.
Field order and wallet order each carry an Autopilot bar. On, and the rows stop being yours to drag: Autopilot tests the sequence against live traffic and keeps the one that converts.
Contact, delivery, payment: whichever order gets the most shoppers to the end.
The method your shoppers actually use goes first, per device.
Switch it off and the rows drag again, exactly where you left them.
When your draft keeps a setting the data argues against, the studio offers a test instead of a lecture: your version on half your shoppers, the recommendation on the other half.
You're keeping 2 trade-offs that cost about 1.8% together. I can test them against the recommended settings on half your shoppers.
Half your shoppers keep seeing exactly what you built. The test runs beside it.
Each trade-off carries its estimated cost, so you can decide which ones are worth a test.
The winner is kept at your confidence line and the studio tells you what changed.
Soft declines clear later in the day. Autopilot retries at the hour that card's bank usually approves, sends a card-update link when a retry won't help, and holds the orders you'd want to look at.
14 payments failed in the last 30 days, $1,212.00 in all. Krepling retried the soft declines and recovered 9; the card-update link brought 3 more back. 2 were flagged high-risk, worth a look before they ship.
A card about to expire gets a link 14 days ahead. A renewal that fails gets three tries over a week, at the hours that bank approves. A shopper about to cancel is offered a month off first.
The moment a case lands, Autopilot pulls the order, the delivery scan and the receipt and writes the response. It waits there, prepared, until you send it, or until a standing approval you set says to.
Seventeen rules across six pages. The ones on by default are safe, reversible housekeeping. The ones that act in your name wait for you.
Type what you want in plain words. Autopilot reads it back as a plan you approve: what changes, who sees it, how long, and what decides it. Nothing runs until you start it.
Every action has a line: what, when, which rule, and an undo where one is possible. Ask Reports what Autopilot did and it answers in a sentence.