kakinang INSIGHTS · 2026-10-08
What should a B2B website ask before preparing a quote?
A visitor who writes “please send a quote” may not yet know which package or specification fits. An AI I want to prepare the fields and human review process for B2B quotations. reception assistant can explain approved public information and collect a usable outline, while a salesperson checks pricing, delivery, and commitments. The purpose of the first exchange is to make the next conversation more productive, not to pretend that a custom offer has already been calculated.
Distinguish the kind of quotation
A quotation request may concern off-the-shelf purchasing, custom manufacturing, recurring supply or a project service. Begin with the intended use and the problem to solve. A manufacturer may need material, tolerance and inspection documents; a consultant may need location, participants and existing systems; a wholesaler needs items, quantities and delivery terms. The public site should say which kinds of work the company accepts and which require individual review. The assistant should answer from the same approved product and service information rather than inventing specifications in conversation.
Use four layers for discovery: the need, the scope, the timetable and the purchasing process. Ask what the buyer is trying to achieve, which products or services and quantities are involved, when a decision and delivery are expected, and who else must assess the offer. HubSpot's qualification material discusses budget, authority, need and timeline as sales considerations. They are a guide for a human conversation, not a command to fire four questions at every first-time visitor. A buyer who is still comparing approaches may benefit more from a clear service explanation than from a premature request for a precise budget.
Make each field useful to a real decision. Quantity can mean one shipment or several batches; a date may mean the requested quote date or the actual delivery deadline. Preserve those distinctions in the lead record. If the customer does not know a condition, mark it for confirmation instead of allowing AI to fill the blank. Examine a small sample of recent quotations that required repeated emails, identify which two or three details were repeatedly missing, and place those questions early in the website flow. This produces a shorter and more useful first exchange than copying an entire procurement form into chat.
Connect reception to an accountable sales process
In kakinang, start with approved public service, product and FAQ knowledge, so website AI can explain scope and the next step. When a visitor expresses concrete interest, lead capture can collect a name, usable contact method, company and need summary. The lead email should reach the people who handle new opportunities. Operational event notifications have a different purpose and should not be treated as the same inbox. Decide who checks unhandled leads and how often. The system can hand over a request, but the team's reply and formal quotation remain human responsibilities.
If there is already an order or customer system, a governed business API can pass permitted fields into the established workflow after configuration. Before activation, specify the origin and validation of each field and how duplicates are handled. If a quantity in chat differs from a later confirmed form, the business must know which value governs. Any action that creates an order, changes its status, or sends terms must follow the company's confirmation policy. The NIST AI risk framework treats governance and documentation as continuing work. For a small team, that can mean recording who approved the answer scope, who owns exceptions and when failures are reviewed.
A usable handoff separates supplied facts from missing details: specification, requested contact route, desired reply date, and unresolved questions. It should not assume that the person asking has purchasing authority or that a large budget makes the opportunity qualified. The salesperson can quote the summary back, check the most important uncertainty, and then propose a technical review, drawing request or standard offer range. If there is no agreed response deadline, the assistant should promise only that the request will be forwarded, not invent a two-hour service commitment.
Test the questions with a hypothetical enquiry
Imagine a display-fixture supplier receiving a Tuesday evening message: “We plan ten new stores by year-end; can you estimate fixture costs?” The assistant can state the published service scope, then ask about intended use, approximate dimensions, quantity per store, installation locations and the first required date. If drawings are not ready, approximate measurements and a contact route are enough for the next conversation. This is a hypothetical workflow, not a customer testimonial or claimed sales result.
The next morning, sales must determine whether “ten stores” is a plan or a confirmed order and whether “year-end” means the first opening or final delivery. Turning ten stores into ten units in an AI summary would distort the estimate; retain the buyer's words and mark the ambiguity. Material, finish, transport and installation may require a specialist. A website can get the parties into the right discussion sooner, but it cannot make technical and financial promises on their behalf.
After release, review a few different cases each week: a complete lead with no follow-up, an incomplete lead that became a useful meeting, an unsupported AI answer, and a repeat visitor. For each, ask what was missing at reception, who owned the next step, and whether the visitor understood the status. Also look at visitors who did not submit contact details without forcing every one into a lead. If people abandon the budget question, make it optional or move it to a human conversation. If sales repeatedly asks for delivery location, move that field earlier. If minimum order size is often asked, have sales approve a public answer and update both the service page and AI knowledge. If requirements are clear but replies are absent, fix ownership rather than adding fields. Sort findings into content gaps, reception friction and handoff failures. This is more informative than treating lead count alone as the outcome.
Questions teams ask
Will asking for a budget at once drive a prospect away?
It can. Ask about use, specifications and timing first, so the reason for a budget question is clear. If price range determines a feasible approach, make the answer optional and explain that it supports an early assessment. Keep a route to a person for visitors who do not know their budget.
Can AI issue a formal quotation?
Only an explicitly configured standard price that can be checked against current information should be shown as such. Custom scope, volume discounts, delivery and contractual terms need an authorised person. AI can organise the request and explain the next step; an estimate must not be presented as a binding offer.
What if the visitor provides only a company name and phone number?
Keep the information that was supplied, ask for the single most useful missing condition, and explain how a salesperson will follow up. Mark unknowns in the handoff. A visitor should not have to complete a long questionnaire to reach a human.
Does connecting an order system make the sale automatic?
No. An API connection needs configured permissions, mapped fields and permitted actions. Creating or changing an order still follows the company's confirmation process. The integration routes approved information; commercial commitments remain governed by business rules.
Sources and further reading
- 101 Sales Qualification Questions — HubSpot
- AI Risk Management Framework Core — NIST