Enquiry Triage Brief for WhatsApp and Email
Version 1.1 · Kramaniti Kosh
Intended outcome
Every incoming enquiry is sorted, answered from approved information or passed to the right person, with nothing sent that a person has not agreed to.
Before you begin
- A sample of recent real enquiries from WhatsApp and email
- The approved answers you already give (price list, stock rules, timings, delivery areas)
- A named person for each type of enquiry that needs a human call
How to use
- Fill in the brief with whoever answers enquiries today. Start from the last few weeks of real messages, not from an ideal list.
- Run it on past enquiries first, in a chat assistant or on paper, and compare its sorting and drafts with what your team actually did.
- Only after that test, use it on live enquiries with every drafted reply approved by a person before sending. Review what it got wrong at the end of each week.
Working template
Role
Sort incoming enquiries from [WhatsApp number or group] and [email address], draft replies for routine questions from approved information only, and pass everything else to the named person with a short summary.
Approved information
List the only sources the assistant may use for replies:
- Price list or catalogue, dated [date]
- Stock and availability rule (for example, “confirm stock with the godown before promising”)
- Business hours, holidays and delivery areas
- Standard replies the team already uses
If a question cannot be answered from these sources, the assistant must say so and hand it over.
Enquiry types
For each type, name what the assistant may do and who receives it.
- Product or availability question. Assistant may draft a reply from the catalogue, but never confirms stock unless the stock rule allows it. Goes to: [Name]
- Price or quotation request. Assistant may share list prices only and flags any request for a discount, bulk rate or credit. Goes to: [Name]
- Order status. Assistant may draft a reply only if the order record is available; otherwise it hands over. Goes to: [Name]
- Complaint, return or damaged goods. Assistant acknowledges receipt only and never admits fault or offers a remedy. Goes to: [Name]
- Payment, credit or dispute. No drafted reply; hand over at once. Goes to: [Name]
- New dealer or supplier. Assistant collects name, city and what they want, then hands over. Goes to: [Name]
- Anything else. Hand over. Goes to: [Name]
Always hand over
- Discounts, credit terms, payment disputes or refunds
- Any promise of a delivery date or time
- Complaints, legal notices or anything that sounds upset or urgent
- Messages the assistant is unsure it has understood, including mixed-language messages it cannot read confidently
- Requests for personal details of other customers or staff
Reply standard
- Language: reply in [language or languages approved by the owner]
- Tone: [for example, polite and brief, as the owner would write]
- Length: short enough to read on a phone
- Never invent a price, stock position, delivery date or policy
Handover note
For every handed-over enquiry, give the person:
- Customer name and channel
- What they are asking, in one line
- Enquiry type and why it was handed over
- Any draft, clearly marked as unapproved
Record
Log each enquiry with date, channel, type, handled by, and whether the draft was sent as written, edited or discarded. Review the log weekly with [Name].
Test before live use
Run at least these past cases: a routine price question, a discount request, a complaint and a message in a language the assistant should not handle alone. Expected behaviour: the first gets a draft, the other three are handed over. Record the result before going live.
Demonstration
This is an illustrative scenario, not a client case study or a measured result. Do not reuse its details as facts about your work.
A sample family-run electrical supplies distributor in Bengaluru receives enquiries from contractors and shop owners on WhatsApp and a shared email inbox. The owner’s son currently answers most of them in the evening.
Sample inputs
Demonstration inputs only: three messages received in one morning. One asks whether a particular switch range is available in white. One asks for a better rate on fifty units for a site. One says a delivered box had two broken fittings. The price list is dated last month; stock is confirmed only by calling the godown.
Example output
Message 1: availability
Type: Product or availability question.
Draft (unapproved): “Thank you for your message. The white finish is in our current catalogue. We will check stock and confirm.”
[Fact within this demonstration] The stock rule does not allow confirming stock without the godown check, so the draft does not promise it.
Message 2: bulk rate
Type: Price or quotation request. Handed over to [Name the owner].
Summary: contractor asks for a reduced rate on fifty units for a site.
Draft (unapproved): “Thank you. [Name] will get back to you about the rate for this quantity.”
Message 3: damaged goods
Type: Complaint. Handed over to [Name the person who handles returns].
Draft (unapproved): “We are sorry to hear this. We have passed your message to [Name], who will call you about it.”
[Recommendation] Ask the customer for a photo only if the owner agrees that this is the usual process.
Handover
All three drafts wait for a person to approve, edit or discard them before anything is sent.
Quality check
Every reply is built only from the approved information; money, complaints and delivery promises always reach a person; nothing is sent without approval during the trial; the weekly log shows which drafts were edited and why.
Limits and human review
This brief does not connect to WhatsApp or email by itself, and it cannot check stock, orders or payments it has not been given. Check your messaging provider’s current rules on automated and business messages, and your obligations for customer data, before connecting any tool. Keep consequential actions with the named human owner.
Edition notes
New in Kosh at version 1.1. Provider-neutral Markdown instructions; no installation or platform compatibility is implied. Runtime behaviour has not been verified across AI providers. Reuse terms have not yet been published. Background reading: Let the Repeated Question Pick the System. Building practical AI support around one workflow, with handover and review rules, is Kramaniti’s Systems Engineering work.