AI for Contracting & Giga-Projects

One view of progress across every site, from the records you already keep.

[object Object]

Delivery across concurrent sites generates more documents, photographs and reports than anyone can read. Orhan reads them: progress and safety from site imagery, variations and claims from correspondence, and a straight answer about where a package actually stands.

Signs you need it

• Progress reporting depends on a site engineer filling in a form, so it is late, optimistic and inconsistent between sites
• Variation orders and claims are argued months later from correspondence nobody has fully read
• Safety observations are recorded after an incident rather than spotted from what the cameras already saw
• Head office and site disagree about percent complete and there is no evidence to settle it

Use cases

• Progress from imagery: Site photography and drone survey read for actual progress against the programme, so percent complete is evidenced rather than asserted.
• Safety and compliance observation: PPE, exclusion zones and housekeeping checked continuously from existing camera feeds, with findings routed to the responsible supervisor.
• Variations, claims and correspondence: Instructions, RFIs and correspondence read and linked to the BOQ item and contract clause they affect, so a claim is assembled from evidence.
• Portfolio-level delivery view: Schedule, cost and risk rolled up across concurrent sites and packages, with the drift flagged while it is still recoverable.

What you get

• Vision models reading your existing site imagery, tuned on your sites rather than a generic dataset
• Findings written back into the tools your team already uses, with the frame or document that produced them attached
• A delivery view that reconciles P6, the cost ledger and what the imagery actually shows
• Exception routing, so a supervisor is notified about their own site instead of everyone receiving a digest
• Model, weights, evaluation data and documentation handed to your team

How it works

• Survey: We look at the imagery and records you already hold and establish what can honestly be read from them before promising anything.
• Baseline: Models are tuned on your own sites and scored against historical periods where the real answer is already known.
• Integrate: Findings are pushed into P6, Aconex or your ERP so they land in the workflow rather than in a separate dashboard nobody opens.
• Operate: A supervised run on one site first, with weekly review of false positives, before it covers the portfolio.

When it is not the right fit

Replacing a structural, safety or quality sign-off by a qualified engineer. The system brings the evidence forward; a competent person still makes the call.

Do we need to install new cameras?

Usually not. Most sites already have CCTV, progress photography or drone survey, and we start by establishing what can be read from what you have.

How accurate is progress read from imagery?

We baseline it against historical periods where the true answer is already known and report the error rate honestly. Where imagery cannot settle a question, the system says so instead of guessing.

Is this monitoring our workers?

It is built for site conditions and compliance, not individual surveillance. Scope, retention and what is and is not observed are agreed in writing before anything is deployed, under PDPL.

Rigel is listed as in design — can we use this today?

The vision product is in design. The delivery view, document and correspondence reading, and forecasting are available now, and we will tell you plainly which part of this is production and which is not.