Refund handling with an approval gate
Refunds stop being a judgement call made in a hurry: the agent assembles the order, the history and the policy, and a person decides with everything in front of them.
The brief
Written like you’d brief a person.
This is the whole setup - a procedure in plain language, with tools and specialist agents referenced inline. When it runs is not written here: that’s the trigger, configured beside it.
Verify the order with #yekar.shopify.get-order, summarise the evidence and the applicable policy in #yekar.slack.send-message, and hand the refund itself to @refund-approver - never issue one without a decision.
How it runs
From trigger to logged outcome.
Trigger
Event - A refund request arrives from support or the store
Verify the order
Confirms the purchase, its state, and any previous refunds.
Assemble the evidence
Collects the thread, the photos, and the policy that applies.
Recommend a decision
Proposes approve, partial, or decline - with the reasoning stated.
Wait for the decision
The refund executes only after an approver decides; silence never becomes consent.
Where people stay in control
Nothing consequential runs on its own.
- Every refund over the threshold waits for an approver.
- The approver sees the exact call that will run, not a paraphrase.
- Undecided requests expire safely instead of executing.
Tools it uses
Runs on accounts you already have.
Shopify
Stripe
Slack
Gmail
Connected once, governed per agent - this use case gets only the tools listed here.
Illustrative savings
$600–$1,500 / month
Assumes 80–250 refund requests a month at 8–12 minutes of investigation each.
Illustrative monthly savings, estimated from typical handling time at public salary benchmarks, reviewed August 2026. Not a promise or a guarantee - your numbers depend on volume and process.
What makes this dependable
Start with the job
See how Yekar.AI would run one of your real processes.
Bring the job, the systems it touches, and the decisions that need a person.