A useful answer starts with the store.
Shopkeeper brings the conversation together with available Shopify context, merchant policy, and approved voice guidance—then replies with the outcome that actually happened.
Support workspace · synthetic thread
Maya Chen · Order #3102
Conversation
Sent after Shopify updated
Context used for this answer
Order
#3102 · Paid · Unfulfilled
Inventory
Small / Sand · 12 available
Store policy
Equal-price swap allowed before fulfillment
Control
Ask first · Approved by merchant
Approved voice guidance
Concise, warm, and specific. Do not over-apologize or promise a ship date.
Representative product composition using fictional details; not a customer transcript.
Resolve the request without separating the answer from the work.
Support quality comes from knowing the current state, respecting the merchant's boundary, and making the final response agree with Shopify.
Answer from context
Use the information available to the conversation instead of filling gaps with a confident guess.
- Order and fulfillment state
- Customer and product context
- Inventory and tracking
- Merchant policies and instructions
Close the loop
Keep the customer response tied to the proposed or completed Shopify outcome.
- Draft or send replies by mode
- Ask before consequential work
- State what actually changed
- Escalate uncertainty or blocked work
Learn deliberately
Improve reusable knowledge and voice with merchant participation, not silent policy invention.
- Save selected policy-gap answers
- Learn from approved edits or examples
- Propose voice updates after enough examples
- Apply a voice change only after approval
Understand first. Act when eligible. Say what happened.
- 01
Read the request
Identify the customer's desired outcome and the information still needed.
- 02
Resolve context
Use available conversation, customer, product, order, inventory, tracking, and policy context.
- 03
Act or ask
Prepare supported work, follow the selected autonomy mode, and pause when judgment is required.
- 04
Reply and record
Send or draft a grounded response and keep the decision and tool outcome reviewable.
Give Shopkeeper reliable sources and a clear boundary.
Connect the store and customer intake, add the policies that matter most, and choose how replies and actions should behave before support starts moving.
- Shopify context for commerce-specific support
- At least one supported customer-intake path
- Store policies and custom instructions
- An autonomy mode, action limits, and an escalation path
Follow the product story.
The practical questions.
- Does Shopkeeper send every reply automatically?
- No. Draft only keeps replies as drafts. Ask first can send routine, structurally safe information replies while consequential work pauses. Trusted is an explicit choice and still respects action limits and approval rules.
- What happens when information is missing?
- Shopkeeper can ask the customer a focused follow-up, ask the merchant for judgment, or escalate. It should not invent a store policy or claim an order changed before Shopify confirms it.
- Does every merchant response become store policy?
- No. A merchant can elect to save an answer to a policy-gap question as reusable knowledge. Ordinary replies do not silently become policy.
- How does voice learning work?
- Approved merchant edits or examples can contribute to a proposed voice update after enough examples. The proposal changes the workspace guidance only after merchant approval.
- Can I export support data?
- Workspace and customer data can be exported as JSON, and action history can be exported as CSV. Export availability and content remain tied to the organization workspace.