Can AI reply to your customers without getting the prices wrong?

If you spend your day writing quotes, AI looks like the obvious answer. It can read an enquiry, understand what the customer wants, and write a polite, well-organised reply in seconds.

The worry is just as obvious. What happens when it gets the price wrong?

That worry is reasonable. The answer is not to avoid AI, but to be clear about which parts of the job it does and which parts it never touches.

Why this matters

A quote is a promise. Once a customer has a figure in writing, a wrong number costs you either the margin or the customer's trust. Most owners would rather keep quoting by hand than take that risk, and they are right to be cautious.

At the same time, slow replies lose sales. An enquiry answered the next morning is often an enquiry that has already gone to a competitor. The point of a well-built AI assistant is to win back that speed without handing over control of your prices.

Where general AI goes wrong with prices

A general chatbot writes whatever sounds most plausible. That is what makes it good at writing and risky at arithmetic and policy. Left to quote on its own, it can:

• round or estimate a figure instead of using the exact one;
• confuse two similar products or sizes;
• apply the wrong tier for the quantity;
• forget delivery, setup or VAT;
• agree to a discount, a deadline or a special deal because the customer asked nicely.

None of these are rare. They are the natural result of asking a writing tool to do a pricing job.

What a safe setup looks like

A safe AI quoting setup splits the work in two. The AI does the reading and the writing. Your own price list, applied by ordinary software, does every calculation. On top of that, a few simple rules keep you in charge.

You approve every email. The assistant drafts the reply and attaches the quote, and both wait for you. You send it as it stands, edit it, or reject it. Nothing reaches a customer that you have not seen.

The AI never sets a price. Once the AI has worked out the product and the quantity, the figures come from your price list, including tiers, extras and delivery. The AI cannot invent a discount, because it is never the source of a number. If a customer asks for a better price, the request comes to you.

Anything shown directly is checked. If you add a chat to your website, it is the one place where AI answers a person without you in between. Every amount it writes should be checked against the calculated price before the visitor sees it, and a reply that fails the check should never be shown.

Running costs are capped. A daily spending limit, and limits per visitor, stop one busy day or one bot from running up the bill. When the limit is reached, the chat simply offers a contact form instead.

Every lead still reaches you. If the assistant cannot write a draft, or a visitor leaves only a phone number, you should be told straight away. The system should never quietly drop an enquiry.

Personal information is handled properly. Under POPIA, visitors should consent before leaving their details, chats that never become enquiries should not keep contact details, and old conversations should be deleted on a schedule.

Questions to ask before you let AI near your quotes

Whoever builds or sells you an AI assistant, ask them plainly:

1. Can it send an email without my approval? The answer you want is no.
2. Where do the prices on a quote come from? The answer you want is your own price list, never the AI.
3. What stops the website chat from quoting the wrong figure? Look for a check on every amount before it is shown.
4. What is the most it can cost me in a day? There should be a cap, and you should set it.
5. What happens to my customers' personal information? Ask about consent, deletion, and who acts as your operator.

Is your pricing ready for this?

The test is simple. If you can write your pricing down as rules, even with tiers, extras and delivery, an assistant can quote from it. If every price depends on a site visit or a judgement call, the AI can still help with the replies, but the pricing stays with you.

See it running

We built exactly this for Taplink, our own smart business card product, and the owner approves every email before it goes out. Read the Taplink case study to see how it works in practice, or see how we build an AI quoting system around your own products and prices.