Serve customers
Support view
Customer messages, suggested replies, evidence chips, customer context, and handoff status in one operator-grade surface.
Operating Model
Every answer follows the same business-safe path
The final product is not a loose chatbot. It starts from reviewed business information, checks whether the answer has enough evidence, records what happened, and gives operators control over anything that can affect money, customer data, or brand trust.
1. Approved business information
Policies, product facts, order rules, service scope, and support boundaries are reviewed before the AI can use them.
2. Evidence-backed response
The assistant uses the approved information it can find, shows where an answer came from, and asks for clarification when evidence is missing.
3. Customer support workflow
The operator sees the customer message, suggested answer, relevant evidence, customer context, and next action in one place.
4. Review and approval
Refunds, account changes, unsupported claims, and sensitive requests pause for a human decision instead of being auto-committed.
Three Product Views
Support, review, and Studio testing work as one system
DuoCode gives merchants three practical views of the same customer-service workflow: live support, internal review, and a testing lane for new service flows before they reach customers.
Explain outcomes
Review view
Inspect what the AI used, why the answer was allowed, what looked risky, and whether the case should be improved or escalated.
Test before launch
Design Studio view
Test new website, support, delivery, and service workflows in a controlled lane before they become customer-facing.
Unified DuoCode cockpit: customer chat, evidence, review, and Design Studio testing in one product surface.
What The Current Product Proves
The product is designed around real support work
The current working product already covers the hard parts that matter in production: trusted business information, customer context, evidence-backed answers, safe actions, and human handoff.
No random web truth
Reviewed business knowledge
Only approved policies, product details, service rules, and workflow notes are used as the basis for customer answers.
Evidence visible
Answers show evidence
Operators can see which approved material supported an answer, so customer replies are easier to trust and review.
Operator control
Actions are risk-aware
Order lookup, ticket creation, checkout, refund review, and escalation are separated by risk so the AI does not quietly overstep.
Review trail
Cases can be reviewed
A manager can go back and understand what happened, what evidence was used, and where a workflow needs improvement.
Business Safety
DuoCode treats risky actions as product workflow, not model confidence
The assistant can draft, explain, and suggest. The product decides what is safe to answer, what needs approval, and what must become a human handoff.
1. Policy answer with evidence
Allowed when approved business information supports the answer
Answer streams with source chips
2. Order or customer lookup
Allowed when the request is read-only and belongs to the right customer or account
Tool event is logged and replayable
3. Refund, cancellation, account mutation
Requires approval or explicit human handoff
Queue item with context, evidence, and decision trail
4. Unsupported or no-evidence request
Clarify, refuse, or hand off based on merchant policy
No unsupported promise reaches the customer
Pilot Packaging
Start narrow, prove the workflow, then scale the system
The launch product is not a generic automation bundle. Each tier adds more reviewed knowledge, operator workflow, and integration depth around the same support base.
Pilot
RM 299/ month
One support workflow, approved-information Q&A, review trail, and a guided Design Studio setup for one store or service lane.
Recommended
Growth
RM 799/ month
Multi-language support, custom policies, source release workflow, audit view, and up to three commerce or internal integrations.
Scale
RM 1,899/ month
Expanded knowledge governance, approval queues, unlimited workflow lanes, priority support, and custom business-system integration.
Private Beta
Bring one real SME workflow and we will map it into the product
The best first pilot is concrete: one product catalog, one refund or delivery policy, one risky action, and one support channel. That is enough to prove whether the system is useful.