ENGINEERING & TENDERS · A GLOBAL ENGINEERING FIRM

BOQ Intelligence

See the price variation on every like-for-like line, across projects and suppliers, so procurement plans tighter and never orders the higher-priced bid.

9 min readJune 2026
  • Private
  • Operationally cost-effective at scale
  • Deterministic & controlled
Your BOQVendor A quoteVendor B quoteVendor C quote
matched on like specs ↓
Concrete C40 · 420 m³Rate
Vendor A412 / m³BEST
Vendor B445 / m³+8%
Vendor C486 / m³+18%

Illustrative example. Same spec, every source, so you buy at the best rate, not the highest bid.

Days → minutes
Per thousand-line comparison

Illustrative

~80–95%
Less review time

Illustrative

1,000s
Line items per tender pack
100%
Lines cited to source
01

Executive summary

BOQ Intelligence reads the bills of quantities, tenders, and specifications that tie up your best engineers, pulls out every line item into a clean structured list, and compares versions, quotes, and specs so deviations surface before they become change orders. Reviews and bids that took days take minutes, and senior engineers spend their time deciding instead of retyping. It runs inside your own environment, every answer links back to the exact file and page, and it stays operationally cost-effective even on the largest tender packs. It is in production today at a global engineering firm.

02

Understanding the problem: what is a bill of quantities?

A bill of quantities, a BOQ, is the detailed, line-by-line list of every material and task a construction or engineering job needs, with the quantity of each and the rate it will be priced at. It is the financial backbone of a tender. It is how a contractor turns a set of drawings and specifications into a number, how a client compares one bidder against another on a like-for-like basis, and how the job is measured and paid once it is under way. A single BOQ for a serious project runs to thousands of lines.

A BOQ never travels alone. It lives inside a tender pack, the full bundle of documents a client issues to invite bids, or that a contractor assembles to submit one. A typical pack contains:

  • The bill of quantities itself, broken into trade or work sections (earthworks, concrete, structural steel, mechanical, electrical, finishes), each with its own run of line items.
  • For every line item: a unique item code or reference, a description of the work or material, the unit of measurement (cubic metres, tonnes, linear metres, "number", "item"), the quantity, the rate, and the extended amount.
  • The technical specifications, the long, dense prose documents that define exactly what each item means: the grade of concrete, the standard a fitting must meet, the tolerance a weld must hold to.
  • Drawings and schedules the BOQ references.
  • Preliminaries, preambles, and the conditions of contract.
  • Addenda and revisions issued during the tender period, which change quantities, add items, or restate a specification after the first version went out.

The pack is created and handled by a chain of people, each of whom adds to it or reads against it:

  • A client or consulting engineer prepares the original BOQ and specification from the design.
  • An estimator on the contractor side reads every line, prices it, and builds the bid.
  • Procurement takes the priced BOQ to suppliers and subcontractors and asks for quotes against it.
  • Suppliers return their own quotes, which never match the BOQ's structure, naming, or units exactly.
  • Project management later measures the actual work against the BOQ to certify payment and manage variations.

Each of those hand-offs is a moment where someone has to read a very long line-item document and line it up against another one: this version of the BOQ against the last revision, the BOQ against a supplier's quote, the priced bill against the technical specification it is supposed to satisfy. And here is the practical reality of the documents themselves: they arrive in every format imaginable. Native spreadsheets from one client, locked PDFs from another, scanned tender packs with handwritten revisions in the margin, specifications in a second language on international jobs. There is no standard.

Doing all of this by hand is slow, expensive, and error-prone, and the cost is mostly hidden:

  • BOQs and tender packs are huge and arrive in every format, including scans. There is no clean digital feed to compare against; someone has to read the document a machine can barely open.
  • Comparing your BOQ to a quote, line by line, takes days. A supplier's quote lists items in a different order, with different descriptions and different units, so matching it back to your BOQ is painstaking manual work across thousands of rows.
  • A missed or mismatched line turns into a surprise cost or a change order later. One item priced against the wrong specification, one quantity that quietly changed between revisions, one scope gap between the BOQ and the spec, none of these announce themselves. They surface on site, months later, as a dispute or a cost the margin did not allow for.
  • Your most experienced people spend that time on data entry instead of judgment. The person reading and re-keying a thousand-line comparison is usually a senior engineer or estimator, exactly the person whose judgment you are paying for, doing exactly the work that needs none of it.

So a senior engineer loses a week to a single comparison, a single overlooked line becomes a dispute on site months later, and the next tender is already waiting. The bottleneck is not skill. It is that the documents are too large and too messy for a person to read carefully and fast at the same time.

The BOQ

Thousands of line items in trade sections, each with code, unit, quantity, rate.

Specifications

Dense prose defining what each line means: grades, standards, tolerances.

Drawings & schedules

The references the BOQ points to.

Addenda & revisions

Mid-tender changes to quantities, items, and specs.

03

What BOQ Intelligence does

BOQ Intelligence reads across all of your engineering, tender, and compliance documents, every BOQ, every revision, every supplier quote, every specification, and turns each one from an unsearchable wall of line items into a clean, comparable, source-backed structure. Where a person can carefully read one document at a time, the agent reads the whole pack and holds every line in a form it can compare against any other. It does not replace the engineer's judgment; it does the reading, extraction, and comparison so the judgment is all that is left.

Across thousands of line items, it produces:

  • A clean, structured table of every line item, code, description, unit, quantity, rate, pulled out of whatever format the document arrived in.
  • A line-by-line comparison between two versions of a BOQ, surfacing every quantity that changed, every item added, and every item dropped between revisions.
  • A reconciliation of a supplier's quote against your BOQ, lining up items that are described differently or ordered differently, and flagging where they diverge on scope, quantity, or rate.
  • A check of the priced bill against the technical specification, surfacing scope gaps where an item in the spec has no line in the BOQ, or a line carries no specification behind it.
  • A short, faithful summary of the long specification text that sits around any given line, so the context of an item is available without reading the whole document.
  • A confidence label on every comparison, marking each match as solid or shaky so a doubtful line is sent to a human rather than waved through.
  • A citation on every answer, linking it back to the exact source file and page it came from.
Line item · like-for-like specVendor AVendor BVendor C
Concrete C40420 m³ · rate / m³412BEST445486+18%
Structural steel S35538 t · rate / t1,1801,090BEST1,240+14%
HVAC duct, galvanised1,240 m · rate / m3841+14%36BEST
Cable tray 300mm950 m · rate / m22BEST25+14%24

Illustrative example. The same line item, matched on spec across every vendor and past project, so the cheapest is obvious and procurement awards each line at its best rate instead of handing the whole pack to one higher-priced bid.

04

Questions it can answer

Anyone, an estimator, a buyer, a project manager, can ask BOQ Intelligence a question in plain English and get a sourced answer back. The kinds of questions it is built to answer:

  • Where does this supplier's quote differ from our BOQ?
  • What is missing, which lines are in our BOQ but absent from the quote, or in the spec but absent from the BOQ?
  • What changed between this revision of the tender and the last one?
  • Which quantities moved, and by how much, between the two versions?
  • Which line items have no rate, or a rate that looks out of line with our previous tenders?
  • How does this rate compare against the same item in our last three tenders?
  • Where are the scope gaps between the bill of quantities and the technical specification?
  • What does the specification actually require for this line item?
  • Which of these comparisons is the agent confident about, and which should a human check?
  • Show me the exact file and page behind this line.
05

How it works

01

It reads the BOQ, tender, and spec documents

In any format, scanned or digital, and it handles other languages where needed.

02

It pulls out every line item

Material, quantity, and rate, all pulled into a clean, structured list.

03

It compares versions, quotes, or specs against each other

And flags the differences.

04

You ask it questions in plain English

Like "where does this quote differ from our BOQ?" or "what's missing?", and every answer links back to the exact file and page.

06

Under the hood (for your technical team)

01

Spec & BOQ Parsing

Reads very long line-item documents into clean, comparable lists.

02

Any-format Extraction

Reads PDFs, scans, emails, spreadsheets, handwriting, images, even video frames, and turns them into clean structured data (OCR included).

03

Document Similarity Matching

Finds duplicate and near-duplicate documents and reused text.

04

Plain-English Q&A

Ask any agent a question in normal words and get a sourced answer.

05

Source Citation & Audit Trail

Links every answer to its exact source, with a tamper-evident log.

The building blocks it's composed from. BOQ Intelligence is assembled from proven capability blocks rather than written as one monolithic model. Each block is a focused, reusable capability that also powers other production agents:

  • Spec & BOQ Parsing reads very long line-item documents into clean, comparable lists. This is the core block; it generalises to tenders, contracts, and catalogues, not only bills of quantities. It is what turns a thousand-row bill into a structured table the rest of the pipeline can reason over.
  • Any-format Extraction reads PDFs, scans, emails, spreadsheets, handwriting, and images and turns them into clean structured data, with OCR included. A scanned tender with handwritten revisions is handled the same as a native digital file, this is the starting point that gets messy inputs into clean structure.
  • Document Similarity Matching finds duplicate and near-duplicate documents and reused text, which is what lets it line a supplier's quote up against your BOQ, across different ordering, naming, and units, and surface where the two diverge.
  • Document Summarisation turns the long specification text into short, faithful summaries, so the context around any one line item is available without reading the whole spec.
  • Confidence Scoring labels every comparison as solid or shaky, so a low-confidence match is flagged for a human rather than acted on as fact.
  • Plain-English Q&A is the front door: anyone asks in normal words and gets a sourced answer back.
  • Source Citation & Audit Trail links every answer to its exact source, with a tamper-evident log.
  • Sovereign Deployment runs the whole agent on your own servers or private cloud, by design, so your tender data never leaves.

The production recipe, turning a tender pack into comparable line items, combines these as: Spec & BOQ Parsing to read the bill, Document Summarisation for context around each line, N-way Document Matching and Document Similarity Matching to compare versions and quotes, Confidence Scoring to label each result, Plain-English Q&A as the interface, all on Sovereign Deployment.

Inputs and modalities. Scanned and digital PDFs, spreadsheets, emails, images, and handwriting, read via OCR; multilingual documents handled where needed. The unit of work is the line item, code, description, unit, quantity, and rate, extracted into a structured table rather than left as free text. That structuring is deliberate: a filled field is far easier to compare and verify than a loose paragraph.

What it integrates with, and how data flows. It reads across all your engineering, tender, and compliance repositories wherever they live, through the Connector Framework that plugs into the tools you already run, Drive, SharePoint, ERP, email, databases, with no rip-and-replace and new sources added in days, not quarters. Documents are ingested, parsed, and indexed entirely inside your environment; questions are answered against that private index, and every answer carries a citation back to the source file and page. The index, the processing, and at the highest tier the model itself all sit on your own infrastructure. Access & Permission Inheritance means access follows the permissions your source systems already enforce: if a person can't see a tender in the source system, they can't see it through the agent either.

07

Built for production

Private

Reads tender, BOQ, and spec documents on your infrastructure; competitive bid data never leaves.

Operationally cost-effective at scale

Cost stays ~flat per pack regardless of size, freeing senior engineers from data entry.

Deterministic & controlled

Pulls every line item into a structured list and links each comparison back to the exact file and page.

Private

BOQ Intelligence runs inside your own environment, on your servers or in your own cloud account on AWS, Google Cloud, or Azure. Your tender packs are commercially sensitive, often under NDA with clients and suppliers, so the documents, the index built from them, and all processing stay on your machine. For the strictest needs the model itself runs on dedicated hardware inside your walls, so nothing leaves at all, the highest level of data sovereignty. Access & Permission Inheritance enforces that access mirrors your source systems, and Source Citation & Audit Trail logs every query, source, and action. That logging is what lets us work to the standards your auditors care about, GDPR, SOC 2 Type 1, and ISO 27001, rather than bolting security on at the end.

Operationally cost-effective at scale

Tender packs are some of the largest documents an enterprise handles, so running cost decides whether this pays back. Instead of pointing a big general-purpose model, billed per token, at thousands of BOQ lines, BOQ Intelligence runs small, job-specific models fine-tuned for parsing and comparing line items on your own infrastructure. A model that only has to read and compare BOQ lines needs far less computing power than a giant model that also has to write poetry, and the hardware is sized to the workload, not to the brochure. The result is a running cost that stays roughly flat as your tender volume grows, rather than scaling with every line read, and that is roughly ten times cheaper in production than the general-purpose approach. We watch that running cost from the first day in production and manage it as volumes grow: FinOps for AI, handled for you.

Deterministic & controlled

A confident wrong number in a bid is worse than no answer at all, so the agent is built for "right, and able to prove it." It fills structured fields, material, quantity, and rate, rather than writing loose paragraphs, which leaves far less room for mistakes to hide and makes every result easy to check against your rules. Confidence Scoring labels every comparison as solid or shaky, so a low-confidence line is flagged for a human rather than waved onto a bid. Source Citation enforces the rule that every answer points back to the exact file and page; if the agent can't find a source, it says so instead of inventing one. No source, no answer. The same BOQ compared the same way returns the same result every time, and an estimator can follow any flagged deviation straight back to the line that produced it.

08

Who benefits

Estimation

Estimators are the people who lose whole days to reading and re-keying thousand-line comparisons today, and they gain the most directly. Instead of working through a bill row by row to price it or check it against a previous version, they get a clean structured table and a line-by-line comparison in minutes. The agent does the reading and the lining-up; the estimator reviews the flagged deviations and the low-confidence lines, and spends the time that frees up on the pricing judgment that actually wins or loses the bid. Their work stops being data entry and becomes review.

Procurement

Procurement lives in the gap between your BOQ and the quotes suppliers send back, quotes that never share your structure, naming, or units. The agent reconciles each quote against the BOQ automatically, lining up items that are described or ordered differently and surfacing where a supplier has quoted on the wrong scope, the wrong quantity, or a rate out of line with your history. Buyers can ask "where does this quote differ from our BOQ?" and get a sourced answer, so they negotiate from a clear picture of the differences rather than discovering a mismatch after the order is placed.

Project management

Project managers feel the cost of the lines that slip through during tendering, the missed item, the quantity that changed between revisions, the scope gap between the BOQ and the spec, because those are exactly what become change orders and disputes on site months later. By catching mismatches and scope gaps before the bid goes out, the agent gives project management fewer surprise costs down the line and a cleaner baseline to measure actual work against. The deviations that used to surface on site now surface at review time, when they are still cheap to fix.

Bid teams

Bid teams work to deadlines that do not move, so the constraint is how much careful checking can be done before the clock runs out. With the reading and comparison handled, bid teams put together faster, more confident bids: they can respond to more tenders inside the same window, and each bid goes out with its deviations flagged and its lines traceable to source. The confidence is not a feeling, it comes from every line being checked and every comparison being labelled solid or shaky before submission.

In short

Read, check, and compare every line of every tender it touches, at the speed of one document and the carefulness of a thousand.

09

Core business value

BOQ Intelligence transforms disconnected bills of quantities, tender revisions, supplier quotes, and technical specifications, arriving in every format, including scans, into a central, searchable, source-backed knowledge and decision-support layer for the whole tendering process. It helps an engineering organisation: cut the days a senior engineer loses to a single comparison down to minutes; catch missed and mismatched lines before they become change orders; reconcile supplier quotes against the BOQ automatically; surface scope gaps between the bill and the specification at review time; compare rates against the history of past tenders; respond to more tenders inside the same deadline window; grow tender volume without growing the estimation team in step; and keep every line traceable to the exact file and page it came from, so the whole bid is auditable.

In simple terms, BOQ Intelligence lets an engineering firm read, check, and compare every line of every tender it touches, at the speed of one document and the carefulness of a thousand.

10

The return (illustrative)

The return on this workflow stacks from four places:

Hours returned

Comparing a thousand-line BOQ against a quote by hand can take a senior engineer two to four days (illustrative); the agent does the reading and comparing, and the engineer reviews the flagged deviations in minutes to a few hours (illustrative).

Error cost avoided

A single missed or mismatched line that surfaces later as a change order can cost a meaningful fraction of project margin (illustrative); structured fields plus citations catch those lines before the bid goes out.

Speed

Faster, more confident bids let you respond to more tenders inside the same deadline window (illustrative).

Scale without headcount

Tender volume can grow without the estimation team growing with it, because the agent's capacity isn't tied to hiring (illustrative).

Time to compare a 1,000-line BOQ against a quote

Illustrative, based on typical workflow mechanics.

By hand, today~2–4 days
With BOQ Intelligence~minutes review
01 day234 days
11

Why teams adopt it

It doesn't change how you write or receive tenders. It reads the painful documents for you and shows you what changed, so work that used to eat days becomes a quick review. There's nothing to rip out and nothing new to learn. It's also a natural first step toward more engineering agents, like bid automation, tender and spec summaries, and document checks.

Start with an assessment.

We scope the right first workflow on your own data and give you an honest go or no-go before you commit to anything bigger.