Construction systems disagree. After three years, no one knows which rates held.
Demiton does not replace your ERP. It does not run your payroll. It does not hold your money.
It reads from your systems, reconciles the result, and holds the record.
What systems does a mid-tier civil firm run?
A mid-tier civil firm runs five.
| Function | System |
|---|---|
| Estimating | Estimating tool |
| Field | Assignar |
| Finance | Business Central |
| Payroll | KeyPay |
| Documents | SharePoint or Dropbox |
Each does its job. The problem is what happens between them.
A cost code gets renamed in Finance after the estimate locks. A field crew records labour against a different category than the estimator planned. A progress claim references a budget version that no longer matches what's in BC. None of these decisions are wrong on their own.
But after three years across forty jobs, no one in the firm can answer which estimator's rates held. Or which crews perform on coastal sand. Or which subcontractors deliver at margin.
The data exists. It's just unstructured against the entities that matter - workers, projects, estimates, ground conditions, vendors.
That's a memory problem. Not a data problem. Not an integration problem.
Why doesn't connecting the systems fix the problem?
Most attempts at this look like sync. Connect the systems. Push data between them. Automate the export.
Sync moves information faster between systems that already disagree. It accelerates drift.
What firms need is a layer that reads from each system, structures the result against the project's entities, and persists the record after the job finishes.
What does the memory layer do?
Demiton reads from your estimating, finance, field, payroll, and document systems through controlled adapters. Every read is bound to identity. Every write is reversible. Every record is sourced.
The structured result is queryable through Claude or ChatGPT. Which estimators' rates held against actuals on coastal piling jobs? gets a sourced answer in five seconds.
After twelve months, the firm knows things it didn't know before:
- Which crews perform on coastal soft ground
- Which subcontractors deliver at margin
- Which estimator's rates held under pressure
- Which conditions drove cost on the last forty jobs
That knowledge does not exist in any of the source systems. It exists in the memory layer. It compounds.
What changes for the firm when the memory layer runs?
Variance becomes visible during execution, not after the project closes.
Estimating gets feedback. Actuals can be compared directly against the original estimate structure.
Reporting reflects what's in the field, not a reconciled summary written three weeks late.
The leadership team stops asking the same questions every project review.
Who is Demiton built for?
We're focused on civil firms doing $50-250M of revenue who already run Business Central, Assignar, KeyPay, and a document system. The memory layer reads from those, structures the result, and feeds it back into how the next job is bid.
For firms not yet ready to integrate live systems, Insights is the entry point: upload your last 24 months of contracts, see where rates held, see where they didn't. Self-serve. $9,000/year.
The longer you run jobs through Demiton, the better the next one gets.
Start with Insights
Upload your historical contracts. See where you priced low. See where you priced right. Talk to it through Claude or ChatGPT.
Diagnose your tender losses.
Upload your historical contracts. Demiton Insights classifies them and shows you where your rates drifted - which jobs ran over, which suppliers held firm, where the margin went. AU$750/month, no per-seat.