Quotes
AI analysis of quotes: what the agent reads, recognises and straightens out before any comparison
The AI analysis of a quote comes before any comparison: the agent recognises every line, attaches it to an item of your taxonomy, makes unit, currency and scope explicit, and takes the context into account. Every value stays linked to where it was read. Without this step, you compare figures that do not talk about the same thing.
The problem
A quote is not data, it is a document
Before comparing or checking a quote, you have to understand it. A spreadsheet filled in by hand understands what the person understood that day, and nothing more: the unit they assumed, the item they thought it was, the option they did not notice on the last page.
Lines that do not look alike
A long description at one supplier, a manufacturer reference at another, an in-house abbreviation at the third. The same item has three names.
Figures without an explicit unit
The same bare price is a price per litre, per can or per pack of six. The quote knows; the spreadsheet cell no longer does.
A context that changes everything
A service on a listed site, a delivery in a difficult area, a request phrased in hours where the quote answers with a lump sum: out of context, the line is unreadable.
The mechanism
What the agent reads, and how it makes it verifiable
Capture reads the quote in its original format (native or scanned PDF, image, spreadsheet, email body) and recognises its structure: header, terms, line table, subtotals, options, annexes. Each line is isolated with its reference, description, quantity, unit, unit price, rebate and amount. A table running over three pages, a shifted column, a skewed scan do not change the result: the line is read, and the place where it was read is kept.
Recognition comes next. The description is attached to an item of your taxonomy and, when it exists, to the article already known in your repository: same product under another reference, same service under another label. The unit is made explicit and, if the tender expects another one, converted. The currency is identified. Options, ancillary charges and services not requested are marked as such. The context of the request (site, constraints, expected quantities) is attached to every line.
Everything the agent decided is visible: the attachment, the conversion, the reason. A line it is unsure about is flagged with its confidence score, never guessed silently. You correct if needed, and the correction is remembered for that supplier. The quote leaves the analysis in a form Compare and Compliance can confront, and that you can re-read.
How an uncertain reading is flagged rather than guessed.
What is recognised
Six things straightened out on every line
The item
The line is classified in your taxonomy: supply, service, maintenance, transport. The attachment and its reason are displayed.
The article
When the repository already knows the product or service under another reference, the link is made, and proposed if doubt exists.
The unit and quantity
Piece, lot, metre, hour, lump sum: the unit is made explicit and the quantity brought to that of the request when conversion is possible.
The currency and language
The quote is read in its language and currency; the rate and conversion date are displayed when a conversion is needed.
The scope
Options, ancillary charges, services not requested, included warranties: each is marked, so as not to be melted into a unit price.
The context
Site concerned, access or safety constraints, original request: attached to the line so that the next comparison takes them into account.
The journey
From intake to usable data
Four steps, traced line by line. The buyer only steps in where the agent flags a doubt.
- 01
Intake and reading
The quote arrives by email, upload or feed. Its structure is recognised and each line is extracted with its position in the document.
- 02
Recognition and classification
Each line is attached to an item, to a known article when one exists, and to the purchase request that triggered it.
- 03
Making things explicit
Units, currencies, scope and context are made explicit. Conversions are made and displayed, never implicit.
- 04
Targeted review
Lines with a low confidence score are presented to you. You approve or correct; the quote is ready to be compared or checked.
What changes
Comparing things that talk about the same thing
Everything downstream depends on this step. A comparison of quotes on badly attached lines produces a false winner; a price check on a misread unit flags gaps that do not exist and misses those that do. AI analysis has no value in itself: it makes the following steps fair, and it makes them verifiable, because every reading decision is displayed and can be challenged by the person who knows the purchase.
For the buyer, the work changes in nature. No more re-keying; they re-read what the agent flagged, and only that. The hundred-line quote and the three-line one take the same review time. And because every correction is remembered, the same supplier is read better next time, without any model being trained on your data and without a template to maintain.
The use case: every quote line linked to the line of the request.
Neighbouring subjects
What this page does not cover
Putting several quotes from one tender on a common grid and quantifying the gaps is the comparison of quotes. Confronting the prices of a quote with history and the price list is quote analysis. Producing a single-format document from heterogeneous quotes is automatic standardization of quotations. This page stops at reading: what the agent understands, and what it shows you to prove it.
The next step: several offers on one grid, item by item.
Capture
Gather and consolidate all your procurement data.
Collection by dedicated mailbox, upload, SFTP, scan, API or ERP export. Reads every format (native PDF, scan, photo, spreadsheet, structured feed), then extracts line by line: supplier, references, quantities, unit prices, discounts, taxes, terms.
The end of re-entry. A procurement history you can finally query.
What Capture detects
- Long documents and multi-page tables
- Unknown layouts
- Several languages and currencies
- Poor scans and handwriting
Nothing is lost. Everything can be checked, everything can be proven.
- The contract clause and the invoice line, highlighted side by side.
- Every extracted value stays linked to the exact place in the document where it was read.
- The same case produces the same decision, today as in six months: the rules are applied deterministically.
- No discrepancy is set aside in silence. Anything that matches no rule is raised, with its reason.
- The agent records what it did, in the order it did it: who, what, how much, when.
Frequently asked questions
What happens when the agent is unsure about a line?
It flags it with a confidence score and the reason for the doubt: unknown description, missing unit, illegible figure. You decide in one click, and the correction is remembered for that supplier. No uncertain value is presented as certain, and none is dropped silently.
Does the analysis work on a supplier never seen before?
Yes. Reading does not depend on a template per supplier: the structure of the document is recognised each time. The attachment to known articles is simply thinner on the first quote, and grows with your corrections, without any model being trained on your data.
Which formats and languages are read?
Native or scanned PDF, image, spreadsheet, email with attachments, feed from a portal. Quotes in several languages and currencies are read in their original language and currency; the necessary conversions are displayed with their rate and date, so that every figure can be traced.
What is the analysis for if I do not compare several quotes?
It makes a single quote checkable and searchable: its lines are attached to your items and articles, its prices can be confronted with the negotiated price and the history, and it joins your repository where it stays findable months later, line by line.
Measurable impact in every environment
More than 5 million procurement documents analysed
Between 1 and 7% of margin recovered
on the scope analysed
From 15 to 45% of time given back to teams, per FTE
depending on the scope and on data maturity
Zylio fits into your existing ecosystem.
The ERP runs the process. Zylio handles the exception and recovers the value that escapes it: invoices without a purchase order, line-by-line price discrepancies, duplicates and overbilling, off-contract spend.
Your data under high security.
Zylio meets the most demanding standards, and nothing is committed without your approval.
- Certifications
- Hosting
- Encryption
- Access
Around AI analysis
- Comparison of quotesThe next step: several offers on one grid, item by item.
- Quote analysisThe prices of one quote confronted with history and price lists.
- Quote OCRThe fields extracted from a quote and the proof of reading.
- Book a demoOne of your quotes read and straightened out live, attachments displayed.
See what this looks like on your own data
Twenty minutes, on a spend category of your choosing. We show you what the agents detect, with the evidence behind it.
- No commitment, on your own data
- Result in 3 weeks
- 20 minutes, no sales pitch
- Your data stays hosted in France
- No change of tool or process