AI Quote Drafting for Small Manufacturers: Buy a Tool, Use Your ERP, or Build Your Own?

Separates drawing-based quoting from catalog RFQ quoting, lists the data a quote draft needs, and compares four routes with the questions to ask each vendor.

Quoting is where AI shows up first

Modern Machine Shop's editor-in-chief put it plainly in June 2026: "AI is quoting jobs in small machine shops across the country." Of all the front-office work in a small manufacturer, quoting gets the most attention from trade press and software vendors, because a slow quote is often a lost job.

But "AI quoting" covers two very different problems, and the right tool depends on which one you have.

Two kinds of quoting

Geometry-driven quotes

You receive a drawing or 3D model and price the job from material, machine time, setups and finishing. This is typical of CNC machining, sheet metal and fabrication. Dedicated quoting software reads the geometry. If that describes most of your quoting, see our page for machine shops and job shops.

Catalog and configured-product quotes

You make standard or configurable products: enclosures, fittings, labels, assemblies built from options. Customers send requests for quote (RFQs) by email listing part numbers, quantities and dates. Pricing comes from price lists, quantity breaks, past quotes and a few rules about rush orders or minimums. A general-purpose AI assistant connected to your documents can draft these well, because the work is reading, looking things up and applying rules, not interpreting geometry.

Many small manufacturers have both kinds. Start with whichever one eats more hours.

What a quote draft needs

Whichever route you choose, the AI can only be as good as what you give it:

Price list and quantity breaks, current, in one place.
Past quotes, ideally marked won or lost.
Lead-time rules: standard lead time by product family, and what counts as rush.
Terms: payment, freight, how long the quote is valid, minimum order.
Margin floors: the lowest price your estimator can accept without asking the owner.
Exceptions: customers with contract pricing, products you no longer quote.

Shops that skip this step get drafts that look confident and are wrong. Writing these rules down is useful even if you never use AI, because it is usually the knowledge that walks out the door when an estimator retires.

Your options, compared

Option	Good fit when	Watch out for
Your ERP's built-in AI	Your system has shipped it (for example, Epicor Prism quote handling)	Test it on your own data; vendor claims are not measurements
Dedicated quoting software	Geometry-driven work, many drawings a week	Subscription cost; may not cover catalog RFQs
Do it yourself with a general AI assistant	One person quotes, low volume, owner likes to tinker	No review queue or audit trail; data handling depends on account settings
Done-for-you build plus training	Catalog or mixed quoting; you want it wired to your inbox and price data with approvals	Setup time; needs clean price data first

On ERP built-ins: Epicor's May 2025 release describes Prism quote handling that "tracks quotes, extracts pricing and terms" and "can suggest a reply," with a vendor figure of "up to 40% less communication." ECI's April 2026 announcement for JobBOSS² did not include quoting AI. If you run enterprise resource planning (ERP) software, ask your vendor what is live today on your version, not what is on the roadmap.

How a quote-drafting workflow runs

An RFQ lands in the quotes inbox.
AI extracts the requested parts, quantities, dates and any special requirements.
It finds the most recent quote for each part and the current list price, applies quantity breaks, and flags anything below your margin floor or outside standard lead time.
It drafts a reply email and a quote document in your format.
Your estimator reviews, adjusts and sends. AI never sets the final price.
Won or lost gets logged, so the next draft has better history to work from.

The time savings come mostly from steps 2 and 3: the looking up, not the writing.

When a do-it-yourself setup is enough

Some owners build their own quote helpers with a chat assistant and a folder of price sheets. That can work well for one person. It starts to strain when:

more than one person quotes and you need consistent prices;
you need a record of what was sent, and why;
customer drawings or export-controlled data are involved (see our rules post);
the price list changes and nobody updates the assistant's copy.

At that point the work is less about AI and more about plumbing: one source of truth for prices, a review queue and a log.

Questions to ask any quoting vendor

Will you run a test on 20 of our past RFQs and show us the drafts next to what we actually sent?
Does it handle our catalog items, our drawings, or both?
Where are our documents stored, and is our data used to train models?
Can the estimator see why a price was suggested?
What does it cost when our volume doubles?
What happens to our data if we cancel?

How quoting fits with order entry

Quotes and orders feed each other. A clean quote history makes order checking easier, because the system has something to compare each incoming purchase order against. If you plan to do both, start by getting quotes into one place in a consistent format. That pays off twice. Order entry is covered in customer purchase orders to sales orders.

Next step

If you are weighing these options, a free conversation from Lunan gives you a short written view of which route fits your quoting and what to clean up first. If AI turns out not to be worth it for your volume, we will say so.