Why can't an AI assistant quote from a price sheet by itself?

Because it predicts text rather than calculating from rules. It can describe a configuration that doesn't exist, apply a discount nobody approved, size a component wrong, or give a different price for the same request on a second try.

How does a quoting engine connect to ChatGPT or Claude?

Through a connector built on the Model Context Protocol (MCP). In ChatGPT it is installed as an app; in Claude it is added as a connector. The assistant calls the engine's tools to look up options, run the checks and price the quote, and shows what the engine returns.

What should I check before trusting an AI-assisted quote?

The same facts should give the same price every time; every number should cite its source; invalid requests should be refused with a reason; prices should come from your own price book; and your rules and quote history should stay in your workspace.

Guide · AI and quoting

Can ChatGPT or Claude quote configured industrial products?

Q: Can ChatGPT or Claude quote configured industrial products?

Not on their own. A chat assistant doesn't know your price book, your options or the codes your product must meet, and it will still produce a confident number. Connected to a quoting engine through a connector, yes: the assistant reads the RFQ and asks the missing questions, the engine sizes, checks and prices from your rules, and the assistant shows the result.

The short answer: not on their own. A chat assistant doesn't know your price book, your options or the codes your product must meet, and it will still give you a confident number. Connected to a quoting engine, yes: the assistant reads the RFQ and asks what's missing, the engine sizes, checks and prices from your rules, and the assistant shows the result with its sources.

What goes wrong when an assistant quotes alone

Paste a price sheet into a chat and ask for a quote, and you'll get something that reads well. The trouble is underneath. A language model predicts text; it doesn't calculate from rules. Sooner or later it will:

Describe a configuration that doesn't exist, because the combination sounded plausible.
Apply a discount nobody approved, or round a price that shouldn't move.
Size a component wrong: a breaker, a conductor gauge, a cooling unit, with no code table behind it.
Give a different answer the second time you ask the same question.

The assistant is good at the conversation. It shouldn't own the number.

How it works with a quoting engine

ChatGPT apps and Claude connectors can call outside tools through the Model Context Protocol (MCP). A quoting engine exposed this way splits the job: the assistant handles the conversation, the engine owns every check and every price.

RFQ, spec or email → Assistant pulls out the facts and asks what's missing → Engine sizes and checks against your rules → Engine prices from your price book → Quote card with sources → Assistant explains it

The assistant never invents a number: if a request can't be built, the engine refuses it and gives the reason, and the assistant passes that on.

What to test before you trust one

Same facts, same price. Ask twice. A rule engine gives the identical quote; a bare model often doesn't.
Every number has a source. The code table, the rating, the catalog page or the price book line.
Bad requests get refused. Ask for something that can't be built and see whether it says no, and why.
Your prices, not the internet's. The quote should come from your price book, not from list prices the model found online.
Your data stays yours. Rules and quote history should live in your own workspace.

Nteli in ChatGPT and Claude

Nteli Quoting is listed in ChatGPT, and connects to Claude as a custom connector. Paste an RFQ, a spec or a customer email and you get back a priced, standards-checked quote in an interactive card, across twenty-three product lines from UL 508A control panels and cable assemblies to data center pods. The same engine runs in the web app if your team doesn't use either assistant. Setup takes a few minutes: use Nteli in Claude or ChatGPT →

Limits, stated plainly: the engine checks the constraints its rules cover and says which standards it didn't check. Your engineer still confirms the design before build; the engine is not a certification body. For how those checks work, see which quoting tools check for engineering errors before pricing.

Ask it for something that can't be built.

Connect Nteli to the assistant you already use, paste a real RFQ, and see what gets checked. Free to use.

Connect it to ChatGPT or Claude → Watch the demos