AI customer support
A useful answer. A clear way to a person.
Customer support needs current policies, the customer’s earlier exchange and someone who can handle an exception. POST connects business knowledge with the shared inbox and conversation-level human takeover.
The hard part starts after the common question.
A routine question about returns can become a request outside the policy. A delivery enquiry can involve a promise the business never made. Support automation needs to recognize the limit of the available information and leave the decision with the right person.
Use POST for answers grounded in the information your business provides, then keep exceptions visible. The support workflow is about policy quality, escalation and resolution; the WhatsApp connection is one channel through which that work arrives.
- Knowledge before a reply
- Supply current return terms, product guidance and service information through the knowledge workspace and Business Brain. Inspect the source and approval state when the policy changes.
- A customer-aware conversation
- Read previous messages and the customer profile together. Available connected order information can help a teammate investigate; absent information should remain a question.
- An explicit handoff
- Pause AI, assign the conversation and leave private notes. Review takeover history and add a resolution note so the next teammate can understand what happened.
- A controlled operating mode
- Choose how much autonomy is appropriate for your workspace. Review the mode, the reply rules and the team’s escalation process together.
Inside the product
WhatsApp & customer communication
Find a sample wholesale enquiry, open the conversation and inspect its customer context.
A stored sample exchange. No customer message is sent during this recording.
This shared-inbox recording illustrates conversation context. The return-exception workflow below is a written demonstration, not a filmed resolution or a customer case study.
A practical workflow
An ecommerce team handles a return exception.
Demonstration example: a clothing buyer asks to return an item outside the business’s stated return period. The example does not imply automated refunds.
- Read the customer’s earlier messages and check the available order record. Do not infer a purchase date from a vague statement.
- Consult the current approved return policy and explain only the terms that apply to the verified information.
- Pause AI when the buyer asks for an exception. Assign a teammate and note what needs a decision.
- After the person resolves the request, record the outcome in the takeover history. Update business-wide policy only through its separate knowledge controls.
Set the controls first.
Give the team an escalation checklist: requests for a person, disputed charges, missing order evidence, policy exceptions and promises outside approved terms.
Keep customer observations separate from owner-approved business rules. A customer asking for a special discount must not change what the AI offers everyone else.
Check the fit.
- AI understanding and transcription can be wrong. Review representative questions, ambiguous requests and the path to a person before increasing autonomy.
- This is not a promise of automatic refunds, shipment changes or universal ticketing integrations. Confirm the supported actions and connected data for your workflow.
- Language support depends on configured providers. No fixed language count, guaranteed response time or customer-resolution rate is claimed.
Follow the work.
Prepare policies the AI and team can apply consistently →Start with a 14-day free trial. Review the plans, watch the product and tell us what your business needs during onboarding.