WhatsAppCustomer conversations
IntercomWeb and in-app chat
ZendeskTickets and hand-offs
ShopifyOrder and stock lookups
HubSpotConversation history
GmailEmail follow-ups
Chat 
agents.Every chat answered.Every request finished.
AI employees that hold the conversation on web chat, in-app and WhatsApp. They answer from approved knowledge, take the action in your systems and write the outcome back.
Trusted by teams at
One conversation, start to finish.
Recorded where you work.
A chat agent picks up the thread, answers, takes the action, hands off what is not its call and writes the outcome back.
WhatsApp · Sarah, E-commerce Support
Answered from approved catalogue knowledgeShopify order lookup- Handled by
- Ishaan, Customer Support
- Customer
- Matched to order
- Shipment status
- Out for delivery
- Tracking link
- Shared in chat
Conversation thread- Started on
- Web chat
- Continued on
- Follow-up
- Email, same thread
- Customer repeated details
- No
Web chat · Sarah, E-commerce SupportTicket · Trail Runner Pro enquiry- Opened
- Answered
- Resolved
- Transcript attached
- Tags applied
Answer
The agent works out what the customer is asking and answers from approved knowledge. If the answer is not there, it says so.
Your knowledge. Your rules. Chat agents only answer from what you approved and only act where you gave access. Everything else goes to your team.
Explore governanceAuthority & handoff
Chat agents finish the request.
People make the decisions.
Starts with
A message on web chat, in-app chat or WhatsApp, or a follow-up your systems queue
Works toward
A finished request: an answer given, an order checked, a document received, a slot booked or a ticket assigned
Leaves to people
Discounts, refunds, advice, disputes and anything outside approved knowledge
Chat agents own
Run the conversations- Answering from approved knowledge
- Looking up orders and accounts
- Collecting documents and details
- Booking appointments and callbacks
- Raising and updating tickets
- Keeping one thread across channels
People keep
Your team decides- Discount and pricing exceptions
- Refund and return approvals
- Coverage and product advice
- Disputes and complaints
- What the agents may answer and do
Handed off with context
- 01Use approved knowledge instead of guessing an answer.
- 02Separate what the customer said from a confirmed system result.
- 03Hand off to a named person with the conversation attached.
- 04Log every conversation with its outcome.
One agent per conversation type
Real employees.
Real conversations on day one.
- Shopify & Zendesk
- Question
- Answered from catalogue
- Stock
- Checked live
- Discount request
- Sent to manager
- Checkout link
- Shared
Pre-purchase questions
Shoppers answered while they decide.
What the agent doesSarah answers sizing and stock questions from your catalogue, shares a checkout link and routes placed-order questions to Ishaan and refunds to Zara.
What you can checkThe chat transcript, the live cart state and the ticket in your helpdesk.
Salesforce & Notion- Setup guide
- Matched to stack
- Blocker
- Assigned to engineer
- Milestone
- Verified
- Credentials shared
- Not accepted in chat
Customer onboarding
Setup moves without manual chasing.
What the agent doesMeera guides new customers through approved setup steps in chat, answers product questions and follows up on blockers until each milestone is verified.
What you can checkThe milestone in your CRM and any blocker assigned to a named engineer.
WhatsApp & Google Drive- Incident details
- Captured
- Documents
- 2 of 3 received
- Reminder
- Sent for missing item
- Coverage decision
- Left to reviewer
Claims intake on WhatsApp
Documents chased, decisions left to people.
What the agent doesAisha captures the incident details, shares the document checklist for the claim and follows up on missing items until the intake packet is ready for review.
What you can checkThe case record, the documents received and the reviewer assigned.
How we set it up
Approve the conversation.
Then let it run.
- 01
Define the conversation
Pick one conversation type. Set the approved knowledge, the actions allowed and the decisions that stay with your team.
Scope gateApproved brief and named business owner - 02
Connect your channels and systems
Connect chat channels, helpdesk, CRM and the systems the agent acts in, with least-privilege access.
Connect gateReviewed access and data-flow map - 03
Test real conversations
Run normal requests and difficult ones: angry customers, refund demands, shared credentials. Check what was said and what was written back.
Evaluate gateAcceptance record and documented open issues - 04
Go live in stages
Start on one channel, like web chat. Review conversations and hand-offs, then add WhatsApp and in-app chat.
Release gateRollout, support and recovery plan
Systems & permissions
Connect the channel and the record.
Chat channelsHold conversations on web chat, in-app chat and WhatsApp.Consent, opt-outs and WhatsApp template approvals.
HelpdeskRaise and update tickets, and hand conversations to a named person.Assignment rules and one owner per ticket.CRMRead the customer record before replying. Write the outcome after.Field-level permissions and write scopes.
Orders and paymentsLook up orders, stock and payment status, and share approved checkout links.Read and write scope, and which offers can be applied.
Documents and calendarsStore documents customers send and book appointments in chat.Storage location, retention and which calendars can be booked.
Team alertsNotify the named owner when a conversation needs a person.Who is notified, and how fast they must respond.Start with one conversation type on one channel. Add more once the first one runs cleanly.
Pilot evidence
Measure the conversations.
Inspect every outcome.
- Helpdesk
- Answered from knowledge
- Tracked
- Actions completed
- Logged
- Tickets resolved
- Verified
Requests finished
How many conversations ended with the request done in your systems, not only an answer given.
- CRM write-back
- Transcript
- Attached
- Outcome field
- Updated
- Next step
- Scheduled
Outcomes recorded
Whether each conversation left a transcript and a clear outcome on the right record.
- Exception queue
- Named owner
- Assigned
- Conversation attached
- Verified
- Time to decision
- Measured
Hand-offs resolved
Whether every hand-off reached its named owner with the conversation, and how quickly it was decided.
Agree the baseline before the pilot, then compare finished requests, recorded outcomes and hand-offs against it.
Buyer guidance
Choose the right kind of chat.
- Use an FAQ bot when…
Customers only need fixed answers and no record needs to change.
- Hire a chat agent when…
Conversations need answers, an action in your systems and a result written back, like order checks, document collection or bookings.
- Keep a person in the chat when…
The customer needs advice, a price exception or a refund decision, or disputes what happened.
- Build a custom workforce when…
Your conversations span unusual steps or legacy systems. Define your own knowledge, actions and completion evidence.
Buyer & implementation questions
What to know before
you put agents in the chat.
Web chat, in-app chat and WhatsApp, with email follow-ups on the same thread. Start with one channel and add the others once the first runs cleanly.
A helpdesk bot stops at an answer. A chat agent answers, then takes the next step in your systems, like checking an order, collecting a document, booking a slot or raising a ticket, and writes the outcome back.
The agent keeps one conversation history across channels, so the customer does not repeat themselves and your team sees the whole thread.
Discounts, refunds, advice, disputes and anything outside approved knowledge go to a named person on your team with the conversation attached. Confirm how hand-offs reach your team during setup.
Chat agents do not ask for or accept secrets in the conversation. Define what to do with sensitive data during setup, and test those cases before go-live.
Sarah handles pre-purchase questions, Ishaan handles order support, Zara handles returns and refunds, Meera handles onboarding, Kavya takes insurance enquiries and Aisha runs claims intake. You can also define your own role.
Customer reviews
From conversations to completed work.
Voice AI for client work
“Using UnleashX gave our team the power to build voice AI agents that deliver real results for our clients. Every use case became faster to deploy and easier to scale.”
Which conversations is your teamnot getting to?
Bring one conversation type. We will map the questions, the systems it touches and the business case.