Business task
Reply faster without losing accuracy
Staff spend time writing similar replies to common enquiries about services, availability and pricing. Slow replies lose customers.
Permitted input
Allowed
- The customer’s enquiry text
- Approved service descriptions and price list
- Opening hours and booking rules
Not allowed
- Payment card or bank details
- Identification numbers
- Health or other sensitive personal information
Workflow
Five steps, one human decision
- Enquiry arrivesA message comes in through WhatsApp or the website and is logged.
- Intent is classifiedThe enquiry is tagged, for example pricing, booking or complaint. Complaints skip drafting and go straight to a person.
- Draft is generatedA draft reply is written using only the approved knowledge base.
- A staff member reviews and decidesThe reviewer checks facts and tone, edits as needed, then approves or rejects. Nothing is sent without this step.
- Reply is sent and loggedThe approved reply is sent and the conversation is recorded in the CRM for follow-up.
Sample
See an example
Customer enquiry · sampleHi, how much is a deep conditioning treatment and can I come on Saturday afternoon?
Show the draft reply
Draft for review · preset example, not live AIHello, thank you for getting in touch. A deep conditioning treatment is KES [price from approved list] and takes about 45 minutes. We have availability on Saturday afternoon between 2pm and 5pm. Would you like us to book a time for you?
The reviewer confirms the price and that the Saturday slot is free before sending.
Limitations
- Drafts can contain errors or out-of-date information, which is why review is mandatory.
- The knowledge base must be kept current by a named owner.
- Sensitive enquiries and complaints should always be handled by a person.
Implementation dependencies
- An approved AI tool with suitable data-handling terms
- A maintained knowledge base of services, prices and policies
- WhatsApp Business API or website chat integration where automation is required
- Staff guidance and training on review responsibilities
Scripted demonstration
This page uses a preset example. It does not process text with AI.