Rules
Refund thresholds, exclusions, escalation paths, final-sale exceptions, fraud signals, and approval limits in one reviewable structure.
For Shopify support teams
Policyfile
Turn refund and return judgment into a policy file an AI support agent can follow. We start hands-on: draft decisions in Slack, learn from every edit, and keep humans approving until the file is trustworthy.
- 2 weeks
- Concierge pilot
- Slack
- Approval loop
- Shopify
- First wedge
The failure mode
Refund rules live everywhere except where agents can use them.
The real policy is scattered across old tickets, Slack threads, manager memory, helpdesk macros, and one-off exceptions. Generic AI support tools can draft replies, but they still have to guess how this merchant actually decides.
How it works
The policy file gets sharper every time a human edits it.
Examples
Real past tickets become decision examples, so the agent can model the judgment behind the reply instead of re-reading stale docs.
Edits
Every approval, rewrite, and rejection is captured append-only, then turned into proposed policy updates for the operator to accept.
Concierge first
We do the work manually before asking you to trust software.
Map the messy policy
Review your current refund policy, macros, and recent edge-case tickets. We define what the agent may draft and what must stay human-approved.
Draft in Slack
Policyfile proposes the customer reply and the operational action. Your team approves, edits, or rejects each one before anything reaches the customer.
Turn edits into policy
Repeated corrections become plain-English rule proposals with provenance, so the next ticket starts from what your team already decided.
What we measure
A pilot should prove operational lift, not vibes.
Approval without edit
Are drafts getting closer each week?
Handle time saved
How much operator and support-agent time moves out of the queue?
Refund leakage caught
Which over-refunds, bad exceptions, or inconsistent decisions did we avoid?
FAQ
Built for trust before autonomy.
Does this send replies automatically?
Not during the pilot. Policyfile drafts the reply and action; your team approves or edits before the customer sees anything.
Which helpdesks do you support first?
The first outreach is focused on Shopify brands using Gorgias or Zendesk, with Slack as the approval surface.
Is this only for refunds?
Refunds are the first workflow because the feedback loop is fast and measurable. The underlying policy-file primitive can extend to other repeated decisions.
What do you need to start?
A short workflow review, your public policy, a few recent edge-case examples, and someone who can approve proposed decisions in Slack.
Now onboarding design partners