A customer sends an Excel inquiry with 60 line items, due the day after tomorrow. Someone in inside sales says: “Can't ChatGPT do this?” The honest answer: partly. A ChatGPT quote works well for structure and text, meaning organizing inquiries, summarizing long texts and drafting cover letters. But ChatGPT and Microsoft Copilot do not know your item master, your prices or your ERP. For short quotes with standard items, that is often enough. For BoQs with many line items, it is not. Many teams already use ChatGPT in sales. The question is when writing a quote with AI really saves time and when specialized AI tools for sales are needed.

Key takeaways

• ChatGPT and Copilot help with structuring inquiries, drafting text and summarizing technical texts.

• They know neither your catalog nor your terms; prices have to come from your ERP, otherwise they risk being invented.

• A chat does not book anything into the ERP, does not produce an auditable GAEB return and does not manage follow-ups.

• In personal ChatGPT accounts, content can be used for training unless you switch that off; business plans do not train by default, according to OpenAI.

• Specialized software pays off when BoQs or long inquiries have to be priced against a large catalog on a regular basis.

Can you create a quote with ChatGPT?

Yes, ChatGPT can draft a quote if you supply the line items, articles and prices yourself, but it cannot reliably work out that content on its own. The model produces plausible text, not verified data from your item master. The more line items and the more technical the inquiry, the greater the review effort.

This is not a weakness of individual vendors but a matter of principle. Microsoft itself states that answers from generative AI are not guaranteed to be 100% correct and that users should review them before sending (Microsoft Learn). For a quote with binding prices, that is the decisive point.

What is ChatGPT good at in B2B quotes?

ChatGPT is strong at anything to do with language and structure. These tasks in the quoting process are well supported: Typical use cases for generative AI in a sales team are cover letters, summaries and pre-sorting inquiries before anyone can create quotes.

• Structuring the inquiry: Turning line items from a free-text email into a table with quantity, unit and description.

• Summarizing long texts: Pulling out the core technical requirements of a line item, such as nominal diameter, pressure rating or protection class.

• Drafting follow-up questions: Spotting unclear or contradictory information and drafting a polite question to the customer.

• Cover letters and accompanying text: Writing the quote letter, notes on lead times or a follow-up email.

• Translating: Understanding inquiries from abroad and writing replies in the customer's language.

With a company plan and uploaded files, this can save noticeable time in individual cases. Important: whatever you upload should be free of data that does not belong in that tool (see privacy below).

Where are the limits of ChatGPT in the quoting process?

The limits are everywhere a quote relies on your own data: catalog, prices, availability, ERP. A general chat has no access to any of it unless you copy the data in, and even then it does not check whether its matching is correct.

Then there is repeatability: the same inquiry can produce different results on two different days. An inside sales team that writes hundreds of quotes a month needs traceable, consistent results. How a BoQ is structured and why long texts matter so much is explained in the article Bill of quantities: structure and processing.

What about privacy and trade secrets?

What matters is which plan you use. In personal ChatGPT accounts (Free, Plus, Pro), OpenAI may use content for training unless you turn off the “Improve the model for everyone” setting (OpenAI Help Center). Personal data and customer CRM data are especially sensitive.

For ChatGPT Business, Enterprise and the API, OpenAI states that inputs and outputs are not used for training by default (OpenAI: Business data). Eligible Enterprise and API customers can also store data in Europe (OpenAI: Data residency in Europe).

According to Microsoft, Microsoft Copilot (formerly Microsoft 365 Copilot) does not use prompts, responses and data from Microsoft Graph to train the foundation models, operates within the EU Data Boundary for EU customers and only shows data the user is allowed to see anyway (Microsoft Learn).

Practical rule: customer data, project prices and purchasing terms do not belong in personal accounts. Ask your data protection officer which plan is approved under which data processing agreement. This article is not legal advice.

Example: A prompt for quote preparation

Good prompts for quotes explicitly forbid the model from inventing prices or articles and require a table of open points. That way you use ChatGPT as a pre-sorter, not as a calculator.

Example (anonymized inquiry, fictitious):

Please quote for school campus project: 40 pcs ball valve DN 25 PN 16 with drain, 120 m stainless steel pipe 28 mm press fitting, 8 pcs balancing valve DN 32 presettable, insulation per GEG. Delivery week 12, free to job site.

The prompt:

Role: You support the inside sales team of a plumbing and HVAC distributor.

Task: Structure the following customer inquiry for quote preparation.

Rules:

1. Do not invent prices, article numbers or manufacturers.

2. Only include line items that appear in the inquiry.

3. Write "UNCLEAR" in every field that cannot be determined unambiguously from the inquiry.

4. List assumptions separately.

Output as a table with the columns Item, Quantity, Unit, Product description, Technical requirements and Open questions for the customer.

Then: a short draft of a follow-up email to the customer covering all open questions.

Inquiry: [Paste inquiry text here, without names, addresses and prices]

For this inquiry, a good answer should ask, among other things, which insulation thickness per GEG is meant and whether “press fitting” requires a specific press profile. What the prompt cannot do: select the matching articles from your range, pull your project prices and create the quote in the ERP. That remains manual work or a job for specialized software.

ChatGPT, Copilot or specialized software: what fits when?

ChatGPT or Copilot is enough if you write few quotes with manageable line items and mainly need help with text. Specialized software pays off when long inquiries or BoQs regularly have to be priced against a large catalog and transferred into the ERP.

For configurable products with fixed rules, CPQ software can also be a fit, see CPQ software or AI. For a market overview, see the article Quoting software compared.

Mercura, an AI software for quote and order automation in technical sales, is built for the second case. Mercura reads inquiries and BoQs in any format (email, PDF, scan, photo, Excel, GAEB), matches each line item to your own catalog, including long texts, checks the specification, sorts suitable options by margin, creates price requests for non-stock items, writes the approved quote into the ERP, exports GAEB or PDF and tracks open quotes. The average match rate is 92%, depending on customer and catalog. Pricing and calculation stay in the ERP, uncertain line items are flagged instead of guessed, and inside sales approves every quote.

On privacy: Mercura is certified to ISO/IEC 27001, GDPR-compliant, hosted in Europe, with a separate tenant per customer; customer data does not train external AI models. At the plumbing and HVAC distributor Sanitär-Heinze, processing time dropped from 2 days to 2 hours per quote (Sanitär-Heinze success story).

To be honest: if you write ten quotes a month with standard items, you do not need specialized software. A company plan for ChatGPT or Copilot plus clean price lists in the ERP is then often the more pragmatic choice. More on the options under Automating quote creation and Quotes and tenders.

Bring a real bill of quantities: in 30 minutes we show you how Mercura handles it in your process. Book a call

Conclusion

ChatGPT and Copilot are good assistants for inside sales: they structure inquiries, summarize long texts and write emails. But a reliable B2B quote is built from your catalog, your prices and your ERP, and a general chat has no dependable access to any of them. Use ChatGPT as a pre-sorter with clear rules and only in the approved company plan. As soon as BoQs and long inquiries arrive regularly, you need a solution with catalog matching, ERP connection and a review process.

Further reading

• Quotes and tenders: turn every request into a quote

• Inside Sales: Tasks and How AI Takes the Load Off

• Quoting Software Compared for Distributors and Manufacturers

• Customer story: Sanitär-Heinze (plumbing and HVAC wholesale)

• Mercura vs. Competitors: AI for Quotes and Orders

Frequently asked questions

Can ChatGPT write a quote? Yes, ChatGPT can write quote texts if you provide line items and prices. Selecting articles from your range and applying your terms is something it cannot do reliably without access to your data.

Can ChatGPT be used for quotes in a GDPR-compliant way? That depends on the plan and your contracts. According to its own statements, OpenAI does not train on your data by default in Business and Enterprise plans; whether use is permissible should be clarified by your data protection officer.

Can Microsoft Copilot access my price lists? Copilot can access content in Microsoft 365 that the respective user is allowed to see, for example a price list in SharePoint. It does not access prices and terms in your ERP without an additional connection.

Can ChatGPT edit a GAEB file? ChatGPT can read and summarize the content of a GAEB XML file, but it is not a GAEB program. An auditable return with unchanged item numbers requires software that supports GAEB, see GAEB format explained.

Which AI tools are there for sales? For sales there are general assistants such as ChatGPT and Copilot, AI features in CRM systems and specialized tools for quotes and order entry. Which one fits depends on whether you mainly process text or mainly data from inquiries and BoQs.

Does ChatGPT invent prices? If ChatGPT is not given prices, it can generate values that sound plausible but are wrong. Forbid this explicitly in the prompt and only take prices from your ERP.