Founding cohort pricing is open: Pro at today's price, locked while you stay. Limited seats. See the offer
@Sam answers your customers from your own policies.
He triages tickets, drafts replies in your brand voice and keeps your FAQ and policies straight. Refunds and first replies wait for your approval.
- 6 agents, one team
- 4 tools, all read-only
- Level 1 by default: Propose
- Every action waits for your approval

Real asks, replayed from our evals
Each card replays a case from the evals we run against the model every week. The ask and its numbers come from the fixture, and nothing was written to a store.
- Scenario 1Simulated
I answer the buyer. You keep the customer.
@Sam owns the conversation with your customers. He drafts replies in your brand voice, triages order issues to the right place, curates the FAQ and the knowledge base, and keeps refund and return policies consistent from one answer to the next. He never quotes a policy without retrieving it from your knowledge base first, cites his source, and escalates the ambiguous cases to a human.

@Sam
Shopify customer support
“Loyal buyers compound.”
- Customer satisfaction
- Response time
- Resolution rate
What @Sam owns, reads and calls
4 tools, all read-only, as classified by the permission matrix that checks every call before it runs.
- Drafts ready-to-send replies in your brand voice.
- Triages order issues: where to look, who to escalate to.
- Curates the FAQ and the knowledge base.
- Keeps refund and return policies consistent across answers.
- Your knowledge base: policies, FAQ and documents.
- Recent store events and your team members.
knowledgeSearchRead-onlylistRecentStoreEventsRead-onlylistOrganizationMembersRead-onlysearchStoreIntelligenceRead-only
You decide how far @Sam goes
@Sam starts at level 1. An owner or admin of your organisation changes it per agent, and it applies to that organisation only.
ProposeLevel 1Default
The agent reads, analyses, and suggests. Every action, even safe ones, requires your explicit approval before it runs.
ApprovedLevel 2
The agent executes safe, whitelisted actions on its own. Destructive, external, or high-spend calls still wait for your approval.
AutonomousLevel 3
The agent runs the playbook end-to-end. Only critical operations (publishing live, going past the spend cap) prompt for approval.
Autonomous businessLevel 4Roadmap
The business runs itself inside the goals you set. This is the third phase of the roadmap, not a setting you can pick today.
Never quotes a policy without retrieving it from your knowledge base.
Cannot issue refunds: a refund waits for your approval.
Escalates ambiguous cases to a human instead of guessing.
@Sam works in a team of 6
You write to @Atlas, who hands each part of your ask to the agent who owns it. Open a profile to see what each one does and will not do.

@Atlas
Shopify AI orchestrator
“You set the direction. I run the room.”
Moves Open @Atlas's profile- Revenue
- LTV
- Profit

@Maya
Shopify marketing
“Lowest CAC. Highest signal.”
Moves Open @Maya's profile- CAC
- ROAS
- LTV

@Marco
Shopify merchandising
“Every product earns its shelf.”
Moves Open @Marco's profile- Conversion rate
- AOV
- Gross margin

@Otis
Shopify operations
“Nothing breaks in silence.”
Moves Open @Otis's profile- Time to live
- Integration incidents
- Automated steps

@Faye
Shopify store intelligence
“Numbers lie. Patterns don't.”
Moves Open @Faye's profile- Insight accuracy
- Detection speed
- Decision impact
Can @Sam change my store without asking me?Autonomy
How do I talk to @Sam?Team
Can I change @Sam's autonomy level?Autonomy
Is there a level above 3?Autonomy
What does it cost to use @Sam?Billing
Why do the scenario cards say Simulated?Scenarios
Put @Sam to work on your store
Describe the job in your own words. @Atlas decides who takes it.