SGTM: quantity extraction at the scale of an infrastructure group

Context

Société Générale des Travaux du Maroc was founded in 1971 and is headquartered in Casablanca. It employs more than 10,000 people and has delivered more than 1,000 projects in transport, energy, water, health and industry, in Morocco and across Africa.

Its references include major port works, Dakhla Atlantique and Nador West Med, and the Grand Stade Hassan II. The group has been listed on the Casablanca Stock Exchange since December 2025.

The challenge: volume, in parallel

A group of this size answers several tenders at the same time, and each one arrives as a large file of drawings and technical documents.

Before a price can be set, the quantities have to be read out of that file: lengths, areas, volumes, counts, the metrics each work package is priced on. Done by hand, that reading takes estimating time away from pricing and from judgement, and the larger the file, the harder it is to be certain that nothing was missed or counted twice.

The challenge is therefore one of scale: keeping the same rigour on every file when the number of files, and the number of pages in each, keeps growing.

Scope of the mission

Mirage deploys agents that extract quantities and metrics automatically from drawings and technical documents, inside the preparation of SGTM's tender responses.

1. Tender file

   Drawings, specifications and technical documents received for a tender.
2. ↓
   Extraction agentMirage

   Reads each sheet and document, extracts the quantities and metrics, and links every figure to the page it came from.
3. ↓
   Quantity schedule

   The extracted quantities are written into a structured schedule, organised the way the estimating team prices.
4. ↓
   Estimator review

   An estimator checks each quantity against its source, then validates, corrects or rejects it.
5. ↓
   Bid figures

   Only validated quantities go on to the pricing of the response.

The mission has two stated objectives: shorten the preparation of tender responses, and make the figures more reliable, because each one can be traced back to the drawing or document it was read from. Neither is a result yet; both are what the deployment will be measured against.

Phased deployment

The deployment follows Mirage's usual sequence: start on the client's own documents, prove the work on a defined scope, then widen it.

• Framing: the document types, measurement rules and work package structure SGTM's estimators use.
• Pilot: extraction on real tender files with one estimating team, every quantity reviewed by an estimator.
• Extension: more document types and more teams, once the pilot's criteria are met.

Mirage's engineers work alongside the estimating teams throughout, and the configuration follows SGTM's own rules rather than a generic model.

Questions about this deployment

• What is Mirage Metrics deploying at SGTM?

  AI agents that extract quantities and metrics from drawings and technical documents, so that SGTM's teams can prepare their responses to tenders faster and on figures they can check against the source.

• Who is SGTM?

  Société Générale des Travaux du Maroc, an infrastructure group founded in 1971 and headquartered in Casablanca, listed on the Casablanca Stock Exchange since December 2025. It employs more than 10,000 people and has delivered more than 1,000 projects in transport, energy, water, health and industry, in Morocco and in Africa.

• Does the agent decide the quantities in a bid?

  No. The agent extracts each quantity and links it to the sheet or document it came from. An estimator reviews it and decides whether it goes into the bid.

• Are there results yet?

  Not yet. The mission has just started. Faster tender preparation and more reliable figures are its objectives, and this page will report what the deployment measures once it does.