For estimators and the people who run estimating

AI for estimators: what it does on a real bid, and what stays with you.

The AI Quoting Engine does the first read of a bid package. It finds every requirement that changes the price, cites the page, flags the conflicts and drafts the quote from your own pricing. You still make every call that needs judgement, and you set the price.

Book a DemoHow we test it →
One public bid, read by the engine
  • 179 ppA waterfront bid package with floating docks and a gangway, read in 44 seconds.
  • 214Statements came back, each tied to the page it came from.
  • p.102The specification names one product for the gangway deck, and the drawing on page 117 calls for grooved timber.

Finding that conflict is the engine's job. Deciding which deck to price, or whether to ask the customer first, is yours.

One public bid, start to finish

On a real bid, most of the hours go into reading.

The Town of Cape Vincent, New York posted its waterfront improvements bid package on its website. It is 179 pages, with floating docks and a gangway, and anyone can download it. We read it page by page, and then we ran it through our own engine.

179
Pages read
44 sec
Reading time
214
Statements, each cited to its page
7
Places the drawings and the specification disagree

About forty of the 179 pages carry the requirements that set the price: the specification around pages 86 to 104, and the drawings around pages 110 to 117. An estimator still has to find those forty pages, and that search is the part the engine takes off your desk.

This was a reading result, not a pricing result. We publish what the engine found, and what it got wrong, on our pilot page, so you can judge it before you trust it.

Video: AI Estimating Software Demo: A 179-Page Bid Package to a Draft Quote (4:07)
What the Cape Vincent package asks for →Our scorecard, including the mistakes →
Who does what on a bid

The engine does the reading. You do the deciding.

Here is how the work splits on a real bid, step by step. For each step, the first line is what the engine does and the second line is what stays with you.

Reading the package

The engine reads every drawing sheet and specification page, and lists the requirements that change the price.

You stop hunting for the forty pages that matter. You read the list, and every line opens its page.

Town of Cape Vincent, New York, 179 pages
Checking the documents

The engine flags the places where the drawings and the specification disagree, and the gaps where nothing is stated.

You decide which reading to price, or whether to ask the customer. The Cape Vincent package has seven such places.

Town of Cape Vincent, New York, 179 pages
Sorting the scope

The engine marks each requirement as yours, the contractor's or another trade's.

You confirm the scope before it goes into the quote, because a scope line is a commercial decision.

How the AI Quoting Engine works
The open decisions

The engine shows the decisions the documents leave open, side by side, with their effect on the price.

In the Cape Vincent demo, grating or timber on the gangway deck is one of them. You choose, and the total follows.

The Cape Vincent demo video
Pricing

The engine drafts the quote from your own past quotes, price lists and cost sheets.

You check the draft and set the price. The margin is never the engine's call.

How the AI Quoting Engine works
Sending

The engine never sends anything to a customer.

Every draft is reviewed by your team, and the decision to submit is a person's.

How the AI Quoting Engine works
What your estimator gets

AI for estimators, as a list you can check line by line.

Four kinds of line land on your desk, and every one carries the page it came from. The examples are from the public documents in our dock guide and on our pilot page.

01

Every requirement that changes the price

Wherever it sits in the package, each requirement comes back with its page, so checking it takes seconds instead of a search.

214On the Cape Vincent package, 214 statements came back, each cited to its page.
02

Where the documents disagree

When the specification and a drawing say different things, both pages are shown. Nothing is quietly decided for you.

p.102, p.117One product named in the specification, and grooved timber on the drawing, for the same gangway deck.
03

Questions for the customer

Where the documents close a door or leave a gap, the sentence is shown, so the question can go out before you price.

p.2Substitute materials or work shall not be permitted. Is an equal allowed to bid at all?
04

The decisions that are yours

The calls the documents leave open come back with their effect on the price, for you to make.

DeckGrating or timber on the gangway deck: both are priced, and the choice is yours.
Then the quote

Then a first draft of the quote, priced from your own past quotes, price lists and cost sheets. Nothing goes to a customer until you have checked it.

What it does not do

It does not set your margin, and it does not decide whether to bid. It does not choose between two readings of the documents; it shows you both. It does not send anything to a customer, and it does not replace the experience that tells you a job is risky.

When it helps, and when it does not

It earns its place on long packages. A two-page request does not need it.

Custom work

Mavlon is the right tool when

  • You quote from bid packages that run to hundreds of pages.
  • Most of your hours go into reading and checking, not into the arithmetic.
  • The requirements for your scope are spread across many sections and drawings.
  • You get more requests than your team can read properly, so some go out late or not at all.
  • Your most experienced estimators are close to retiring, and what they know sits in their heads and in old quotes.
Short, clean requests

You do not need it when

  • Your requests are two to four pages and name the product plainly.
  • You quote parts from CAD models. A quoting tool that reads geometry fits that better.
  • You sell from a catalogue by model number, and a configurator already prices it.
  • Your slow step is approvals or the sales process, not reading the documents.

We make Mavlon, so weigh this sort with that in mind. If most of your requests are short and clean, you do not need this, and we would rather say so now. Corrections are welcome at atishay@mavlon.co.

Do not take our word for it

How an estimator should test any AI for estimators.

Give any vendor, including us, a package your team has already quoted, and judge the result the way you would judge a new hire.

  1. Does it read every page, or only the section you point it to? On a public 668-page headworks book, the requirements for one machine sat in fourteen sections.
  2. Does every requirement come back with its page, so you can check it in seconds?
  3. When two documents disagree, does it show both pages, or quietly pick one?
  4. Does it turn gaps into questions for the customer, or fill them with guesses?
  5. Does the draft use your own rates and past quotes, and can you see where every number came from?

Did it find the item? Did it cite the page? Did it show you the conflict, or quietly pick a side?

Download the Cape Vincent package →Our pilot protocol and pass bar →
What we have published so far
  • Cape Vincent, New York waterfront package, 179 pages: read in 44 seconds, 214 statements each cited to its page, and 7 places where the drawings and the specification disagree.
  • Hailey, Idaho headworks book, 668 pages: the 14 sections relevant to the equipment, found without being told where to look, and their requirements pulled with page references in 102 seconds.
  • Odessa, Missouri bar screen, 4 pages: the sentence on page 2 that says substitutes are not permitted.
  • These are reading results. The pricing column on our pilot page is nearly empty, and the page explains why.
The pass bar: seven of ten totals within 15 percent, agreed in writing first.
We did the reading first

Guides for estimators who quote from bid packages.

GuideReading bid packagesWhat five real public packages taught us.GuideHow to capture estimator knowledgeBefore it walks out of the door.GuideQuote turnaround timeWhere the days actually go.GuideUsing ChatGPT to read bid packagesWhat actually breaks on real packages.GuideAI specification reviewThe reading on its own, before anyone prices.ToolSpecification compliance matrixAn Excel template for answering a specification line by line.
What is AI for estimators?

It is software that does the first read of a bid package for an estimator. Mavlon's AI Quoting Engine reads the drawings, the specification and the addenda, lists every requirement that changes the price with its page, flags where the documents disagree, and drafts a first quote from your own pricing. The estimator checks it and sets the price.

Will AI replace estimators?

No. The AI Quoting Engine does the first read: it finds the requirements, cites the pages and drafts the quote. Your estimator checks the draft, makes the judgement calls and sets the price. The point is that the same team can answer more requests, and spend its time on decisions instead of on reading.

What is the best AI for estimators?

It depends on what your requests are made of. If they are CAD files of parts, a quoting tool that reads geometry fits. If you sell configured products from a catalogue, a configurator fits. If they are bid packages with drawings and specifications, you need an engine that reads the whole package and cites every page, which is what Mavlon builds. Our buyer guide compares the tools side by side. Read the AI quoting software buyer guide.

Can we not just use ChatGPT for this?

For a short, clean request, a chat assistant can help. On a full bid package it breaks in known places: it cannot hold hundreds of pages at once, it does not tie each requirement to its page, its reading changes from one run to the next, and it fills gaps with guesses instead of flagging them. We wrote up what breaks on real packages. Read what breaks when ChatGPT reads a bid package.

What does it need from us to get started?

Once, at setup, it needs some past bid packages with the quotes you filed for them, and your pricing workbook with its rules and rates. After that it needs only the bid package for each new request. An NDA is signed before any document of yours moves.

How do we test it on our own bids?

Bring quotes you have already priced. The pilot is a blind test on your own past quotes, scored against a pass bar that is agreed in writing before anything runs: seven of ten totals within 15 percent. The whole protocol is published on our pilot page. Read the pilot protocol.

Who sees our customers' documents?

An NDA is signed before any document of yours moves, and your documents are never used to train AI models. Deployment options and encryption are set out on our security page. Read the security page.

Questions estimators ask

Asked plainly, answered in writing.

Bring one bid package you have already quoted.

We read it together on a call. Compare what the engine finds with what your team found. If you would rather not share a document on day one, we can read a public package together instead. An NDA is signed before anything of yours moves.

Book a Demo