Skip to content
Demiton
Demiton
Agents

Agents that show their working.

An estimator spends two days reading a tender before anyone can say whether it is worth bidding. A quantity surveyor takes a client schedule of rates on trust because re-measuring it from the drawings costs more than the error usually does.

Both of those are reading jobs. Demiton agents do the reading and hand back something you can check - every claim cited to the record it came from.

Should we bid this one?

Tender Brief

Composes the tender launch brief: a two-pager telling an estimator or GM everything Civil Memory knows about a tender and how your firm is positioned to pursue it. Scope and quantities, compliance posture, track record against the buyer, risks, and a pursue recommendation.

Scope and quantities drawn from the tender documents
Your compliance and prequalification posture for this buyer
Track record: what you have bid and won against this agency before
Risks, and a pursue or decline recommendation

Every claim in the brief cites the fact or document it came from. Where Civil Memory does not know something, the brief says so rather than filling the gap.

Do the quantities actually add up?

Schedule of Rates Verification

Takes a client schedule of rates and independently re-derives each verifiable quantity from the drawings, then reconciles its own derivation against the schedule and reports where the two disagree.

Quantities re-derived from drawing tables and symbol counts
Deterministic geometric measurement and spacing arithmetic
Bar-schedule tonnage checked against the schedule
A discrepancy report with the visual evidence behind each finding

The measurement is deterministic, not estimated. Each discrepancy carries the crop of the drawing it was derived from, so a quantity surveyor can check the working rather than trust the output.

How they run

They write to the record, not to a chat window

An agent run produces facts in Civil Memory - sourced, dated, attributed to the connector that fetched them. The output is a durable record you can query later, not a message that scrolls away.

Every run is identity-bound and audited

Agents execute under a named user through Microsoft Entra ID, not a shared service account. Every step, tool call and document read is logged. You can see exactly what the agent looked at before it reached a conclusion.

They stop rather than guess

Agents run against a budget of steps, tokens and wall-clock time, and they are built to report a gap rather than invent a figure to fill it. A brief that cannot find a quantity says the harvest is missing.

Agents read what your team does not have time to.

Agents read the tender and project documents held in your connected systems, so they run on a Connected deployment with document sources in place. Tell us which documents you hold and where, and we will tell you which agents earn their keep.

Talk to us about agents