Retail and e-commerce
Every order question answered, every return handled.
- Order status and delivery tracking
- Address changes before dispatch
- Returns, exchanges, and replacements within policy
Product
AIPranata gives customer service, operations, and compliance teams one platform to design AI concierges, connect them to the systems where work happens, control exactly what they may do, and review every conversation. Six modules cover the full lifecycle, from the first greeting to the audit record.
Trusted by teams at





01 · Concierge Studio
Concierge Studio is where you decide who your concierge is. Customer service leads set the persona, tone, greeting, and languages for each concierge, then choose the knowledge it may draw on, from returns policies and delivery FAQs to product catalogues.
Every change is saved as a new version. Compare versions side by side, try a draft against past conversations before customers see it, and roll back in one step if something does not behave as expected.
Concierge settings
Version 18 · published
Persona and tone
Name, voice, greeting, and level of formality, set for each concierge and brand.
Languages
Replies in the customer’s language, with wording tailored per language where needed.
Knowledge sources
Policies, FAQs, and product catalogues, kept in sync on a schedule you choose.
Draft and test
Run a draft against past conversations before it reaches customers.
Versions and rollback
Every published change is recorded, comparable, and reversible.
Channel settings
Separate greetings, hours, and escalation contacts for web chat and email.
02 · Actions and integrations
Actions turn a concierge from something that answers into something that resolves. Each action, such as looking up an order, creating a replacement, or rescheduling a booking, is defined against one of your systems with an explicit scope.
Connect order management, CRM, booking, helpdesk, and payment systems, or describe your own APIs as custom actions. Parameters are checked before any call reaches your systems, and every request and response is logged against the conversation.
Actions
| Action | System | Scope |
|---|---|---|
| lookup_order | Order management | Read |
| create_replacement | Order management | Write · within policy |
| issue_refund | Payments | Write · approval above $150 |
| reschedule_booking | Booking | Write |
| update_address | CRM | Write · verified customers |
System connections
Order management, CRM, booking, helpdesk, and payment systems.
Custom actions
Expose your own API endpoints as actions with defined inputs and outputs.
Explicit scopes
Read or write access per action, limited to verified customers where needed.
Parameter checks
Structured inputs are validated before a call reaches your systems.
Call logging
Every request and response is recorded against the conversation.
Revocable access
Pause a single action or disconnect a system at any time.
03 · Policies and approvals
Policies define what a concierge may do on its own. Rules can depend on the amount involved, the customer’s tier, the type of request, the channel, or the time of day, and every proposed action is checked against them before it runs.
When an action falls outside a rule, it becomes an approval request. Approvers see the customer, the order, the concierge’s reasoning, and the policy that triggered the request, then approve, decline, or adjust it. The concierge keeps the customer informed while they wait.
Approval request
Refund $240 for Maria K.
Exceeds automatic refund limit of $150
Rule conditions
Amount, customer tier, request type, channel, and time of day.
Approval routing
Send requests to a team or named approver, with a backup if no one responds.
Full context
Customer, order, concierge reasoning, and triggering policy in one view.
Approve, decline, or adjust
Change the amount or the action before it runs.
Customer updates
The concierge tells the customer what is happening while they wait.
Decision records
Every approval is stored with the policy and the approver.
04 · Handoff workspace
Some conversations need a person. When a customer is upset, asks for a human, or needs something outside policy, the concierge hands the conversation to the right team with everything they need to resolve it without starting over.
Agents receive a summary, the customer’s history, the actions already taken, and a suggested next step. They can reply directly, take actions themselves, or return the conversation to the concierge once the exception is handled.
Handoff
Summary
Customer disputes a $240 refund amount after a damaged delivery. Photos received. Refund above automatic limit.
Already done
Suggested next step
Approve full refund and offer express replacement.
Handoff triggers
Customer requests, frustration, policy exceptions, VIP status, or sensitive topics.
Routing
To teams, queues, or named owners such as account managers.
Summaries
What the customer needs, what was tried, and what to do next.
Shared history
Previous conversations, orders, and preferences in one panel.
Return to concierge
Hand routine follow-up back once the exception is resolved.
Helpdesk sync
Work from your existing helpdesk if your team prefers.
05 · Analytics and quality
Analytics show what concierges resolve on their own, what they hand off, and why. Trends are broken down by topic, channel, language, and concierge, so you can see where to improve knowledge, adjust policies, or add actions.
A review queue collects conversations that deserve a second look, such as low-confidence answers, requests for a person, or approved policy exceptions. Reviewers rate replies and leave notes, and recurring gaps become suggestions for new knowledge or actions.
Quality review
Flagged for review
Top knowledge gaps
Resolution trends
Resolved, handed off, and abandoned conversations over time.
Breakdowns
By request type, channel, language, and concierge.
Review queue
Low-confidence answers, requests for a person, and policy exceptions.
Reply ratings
Score replies and leave notes for the team.
Knowledge gaps
Recurring questions the concierge could not answer well.
Customer feedback
Post-conversation ratings next to each transcript.
06 · Audit log
The audit log records everything concierges and people do in the platform: every message, lookup, change, approval, configuration edit, and permission change, with who or what performed it and which policy allowed it.
Filter by customer, concierge, policy, or person, follow a single case from the first message to resolution, and export records for internal review, customer disputes, or audits.
Audit log
| Time | Actor | Event | Policy |
|---|---|---|---|
| 14:02:11 | Concierge · Ava | Order #48213 retrieved | n/a |
| 14:02:12 | Concierge · Ava | Replacement eligibility checked | RPL-02 |
| 14:02:40 | Concierge · Ava | Express replacement created | RPL-02 |
| 14:31:05 | Rina S. · Operations | Refund $240 approved | RFD-03 |
Complete coverage
Messages, actions, approvals, configuration, and permission changes.
Actor and policy
Who or what acted, and the rule that allowed it.
Case timelines
Follow one conversation from first message to resolution.
Filters and search
By customer, concierge, policy, person, or date.
Export
Download records for reviews, disputes, and audits.
Retention settings
Keep records for as long as your policies require.
Platform
Use cases
The same platform serves retail, hospitality, financial services, and telecommunications, with rules that fit each industry.
Retail and e-commerce
Hospitality and travel
Financial services
Telecommunications and utilities
Pricing
Choose the package that fits your team. Both include the concierge platform, policy rules, human handoff, and a complete audit log.
For teams launching their first concierge on web chat.
Rp 500,000/month
For teams running concierges across channels, with approvals and quality review.
Rp 1,500,000/month
Prices are in Indonesian rupiah, exclude VAT, and are paid monthly in advance. Need more volume or custom terms? Contact us
Sign in to create a workspace, connect a system, and configure your first concierge.