AI use case · Construction
AI first-pass review for submittals and RFIs, with the consultant making every decision.
For consultants, contractors and document controllers whose reviewers spend more time finding the right clause than judging the submittal. AI gathers the references and drafts a checklist. The engineer decides.
What is AI first-pass submittal review?
AI first-pass review reads an incoming submittal or RFI, finds the specification sections, drawings and earlier correspondence it relates to, and drafts a checklist of compliance points and missing information. Every point links to its source page. The reviewing engineer checks the draft and the references, makes the decision and records it in the project system.
The problem
Reviewers search before they can judge.
On a UAE building project, one material submittal can carry data sheets, test certificates, a sample schedule and a cover letter, sometimes part Arabic and part English. Before the engineer can say yes or no, someone has to find the specification section, the drawing references and any earlier comments on the same item.
RFIs bring the same search. The question is often short, but the answer depends on a clause, a detail or an earlier instruction somewhere in the project file. When the review queue grows, contractors chase for responses and the review periods in the contract keep running.
Most of that search is clerical work. The judgement is not, and it should stay with the consultant.
Clause hunting
Reviewers open the specification, the drawing register and the transmittal log to find what applies before they read the submittal itself.
Missing documents found late
An incomplete submittal is often spotted on the second pass, after it has already waited in the queue.
Uneven comments
Two reviewers can return different comments on similar items, because each one starts the search from scratch.
How it works
One submittal, from transmittal to recorded decision.
An illustrative workflow. The document system, review checklist and approval codes are agreed with your team for each project.
Arrival
A submittal or RFI arrives in your document system, such as ArkanCS or the register you use today, with its transmittal and attachments.
Reference search
AI finds the specification sections, drawing numbers and earlier submittals or RFIs on the same item, and lists them with page references.
Draft checklist
It drafts the compliance points to check and flags missing items, such as a test certificate the specification asks for. Each point links to its source.
Engineer review
The reviewing engineer opens the draft beside the documents, confirms or rejects each point and adds their own comments.
Decision and code
The engineer chooses the approval code and signs off. AI never assigns a status.
Recorded in the project system
The decision, comments and reviewer are recorded against the submittal or RFI, and the AI draft is kept with them for audit.
Who does what
What the AI does · What a person checks.
What the AI does
Reads submittals, RFIs and attachments in Arabic and English
Finds the related specification sections and drawing references
Lists earlier submittals, RFIs and comments on the same item
Drafts a checklist of compliance points, each linked to its source
Flags documents the specification asks for that are missing
What a person checks
Every compliance point and every source reference in the draft
Technical and design judgement, including whether an alternative product is equal
The approval code and the comments returned to the contractor
Anything with safety, authority or contractual consequences
The final sign-off recorded in the project system
First pilot
A first pilot on one package of real submittals.
We test on your own project documents before anyone talks about results. Scope, timeline and price are confirmed in a written proposal.
One discipline
Start with one trade, such as MEP material submittals, where the specification sections are clearly defined.
One document source
One document system or register, and one agreed way to read submittals and return the draft to the reviewer.
Closed submittals as the test set
Past submittals with known decisions and comments, split into a development set and a held-out set.
Agreed acceptance
Agree with your reviewers what a useful reference and a correct missing-item flag look like, before the evaluation runs.
Data, hosting and Arabic
Where project documents go, and how Arabic is handled.
Hosting
The application and stored data can be hosted on Azure UAE North, on private servers managed by Vloud, or in your own cloud account. The AI model is used through a provider agreed per project, and its processing region may be outside the UAE.
Access
Access follows project roles: a reviewer sees the packages for their projects and disciplines, nothing more.
Arabic
Arabic cover letters, authority comments and mixed-language documents are tested as their own sample, not assumed to behave like English ones.
Drawings
Text in title blocks and notes is read where the file allows it. Scanned or image-only drawings are flagged for a person, not guessed.
Honest limits
Where it does not work yet.
- AI prepares the review and the consultant decides. Design judgement, approval codes and sign-off are never automated.
- We do not quote accuracy or time savings before testing on your submittals. Results are measured on a held-out sample first.
- If the specification or drawing register is incomplete or out of date, the references will be too. The workflow does not fill the gaps.
- Reading dimensions, symbols and details inside drawings is limited. Drawing checks stay with the engineer.
- For an RFI, the workflow gathers references and can draft a reply for the engineer to edit. It never sends one.
What drives cost and timeline. The number of disciplines and document types, the size and structure of your specifications, drawing formats and scan quality, languages, the document system's integration route and hosting requirements. We don't publish a price list; scope, timeline and price come in a written proposal after the discovery call.
Evidence
What we can show you.
Every example is labelled: delivered work, our own product, a demonstrator, or an illustration with synthetic data.
- Delivered work
ArkanCS at Al Mukab AlFudhi
Al Mukab AlFudhi Engineering Consultancy runs a G+5P+15 building project on ArkanCS, with submittals, drawings and inspections in use. It shows the review records a first pass would feed. No AI review is claimed on this project, and RFIs are not among its confirmed modules.
- Our product
ArkanCS
Our own construction management platform for inspections, submittals, drawings, BOQs and payments. A first-pass review can connect to ArkanCS or to the document system you already use.
- Illustrative
How to test AI document extraction
The evaluation method we apply in a pilot, with a development set and a held-out set. The same approach is used to check whether the references a first pass finds are the right ones.
Before you start
Questions buyers ask.
Does it replace the reviewing engineer?
No. AI gathers references and drafts a checklist. The engineer checks each point, makes the technical judgement and assigns the approval code. The decision and the reviewer's name are what is recorded.
Which documents does it read?
Submittals, RFIs, transmittals, data sheets, test certificates, specifications and the text in drawing title blocks and notes where the file allows it. The document types in scope are agreed for the pilot.
Does it work with ArkanCS or our existing system?
It can work with ArkanCS or with the document system you use today, if that system has an API or an agreed export route. We review it before confirming the integration scope.
Can it handle Arabic correspondence?
Yes, Arabic, English and mixed documents. Arabic cover letters and authority comments are tested as their own sample during the pilot, because layout and scan quality vary.
Do you understand how consultants review submittals?
Our founder is a mechanical engineer with a master's in construction management and experience across MEP and tendering. We also build ArkanCS, so the workflow follows how review decisions are recorded on a project.
Where is the data hosted?
The application and stored data can be hosted on Azure UAE North, on private servers managed by Vloud, or in your own cloud account. The AI model is used through a provider agreed per project, and its processing region may be outside the UAE.
Who owns the code?
You own the source code and IP of the custom workflow we build for you. Our own products stay ours, and any licence terms are agreed separately.
Related
Related AI use cases.
Tender and BOQ document analysis
Read tender packs and BOQs into structured items, flag missing scope, unusual conditions and key dates, and keep every pricing decision with your estimator.Daily site report and meeting minutes summarisation
Turn site notes, voice memos, photos and meeting recordings into structured daily reports and action lists that your site engineer checks before issue.Contract clause and variation tracking
Find notice periods, variation clauses and obligations in construction contracts, with each answer linked to its clause, and track every variation to closure.ArkanCS
Our construction management platform for submittals, drawings and inspections.AI document processing
The service page for document reading, checks and review.Construction software in the UAE
How we build and run project systems for contractors and consultants.
Next step
Test it on your own documents.
Tell us what the work looks like today. We reply with the questions we need answered before a pilot can be scoped.
