AI for construction companies
The site moves only when the paperwork does.
Margin in construction rarely disappears in one place. It leaks through work performed before it is priced, quotes compared by eye, a DURC di congruità computed the week it is due, and a progress statement assembled from photos and memory. Every one of those is a queue that a site manager, an estimator or the office absorbs in the evenings. The queue is where an agent belongs, with the person who knows the site keeping the decision.
- Built on top of the estimating and accounting software you already run
- A named person signs before anything reaches a client, a supplier or an authority
- Most of the six queues below need code and a rule, not a model
Six queues, six roles
Put an agent on the queue, not on the org chart.
Each role owns one recurring queue. It drafts, it shows its evidence, and a named person signs before anything consequential moves. The last column is the sector’s AI Boundary Map: whether the step is mostly a rule, mostly a model, or stays with a person.
| Role | What it drafts | Who signs | What it reads | Mostly |
|---|---|---|---|---|
| Estimating desk | A bill of quantities from the drawings, priced against the regional price list, with every item that had no match flagged. | The estimator, item by item, before the quote goes out. | Drawings, the regional prezzario, your own historical quotes. | a model |
| DURC di congruità check | The labour-cost congruity calculation per site against the CNCE thresholds, updated every time payroll or the contract value changes, with the gap to the threshold shown in days of labour. | The office, before submission. | Payroll, the contract, the Cassa Edile denuncia. | code and rules |
| Progress statement (SAL) clerk | The progress statement from site measurements and delivery notes, with each quantity traced to the document that supports it. | The site manager and the works director. | Site measurements, DDTs, the computo. | code and rules |
| Supplier quote comparison | Three or five supplier quotes normalised to one line list, with missing items, unit mismatches and non-standard conditions flagged. | The buyer. | Supplier PDFs and emails, your RdO. | a model |
| Cassa Edile monthly filing | The monthly denuncia from recorded hours per worker per site, with anomalies against the previous month listed. | The payroll consultant. | Timesheets, payroll, the previous filing. | code and rules |
| Subcontractor compliance | The document set per subcontractor per site, with expiries thirty days out and the contract drafted from the accepted quote. | The office, and the site manager for site access. | PEC, the document store, the DURC portal. | code and rules |
The determinism ratio
How much of this is actually AI.
Four of these six queues are rules and lookups: the congruity threshold is a published table, the SAL is arithmetic over documents, the denuncia is a transformation of hours, the compliance set is a checklist with dates. Only the two that read unstructured documents, drawings and supplier PDFs, need a model, and even there the model extracts and a rule prices. That is the shape the AI Boundary Map usually takes in this sector, and it is why the running cost stays low and every number can be shown to a works director as a rule rather than argued for as a sample.
4 / 6 queues mostly rules · 2 / 6 with a model
ProofAll engagement data, with sources →
No client names. Real patterns.
Every engagement produces the same five documents, whatever the sector. What we can show is anonymised: the pattern, the number, and where it applies.
Across the workflows SUPALABS has mapped, the step that needs a model is the minority: in one European manufacturer’s order handling, three of eleven. Construction estimating follows the same shape, with extraction in the model and pricing in a rule (SUPALABS engagement data, 2024–2026).
Related case studies
From the Official Price List to the Client Quote, Without Losing the CodeAn Italian construction estimate has to cite published price-list codes. Copy those lines into a spreadsheet and the link to the source is gone — which is exactly when someone asks where a number came from. Read the case study →
Material Decisions Get Locked in Phase 3. The Data Arrives in Phase 5.Life-cycle assessment tells an architect what a building will cost the climate — after the materials have been chosen. We built the decision support that arrives while the choice is still open. Read the case study →
An Inbound Sales Agent That Qualifies Leads in Natural ItalianConstruction professionals ask about pricing at eleven at night. We built a messaging agent that answers in natural Italian, qualifies through five questions, separates enterprise from consumer, and books the demo. Read the case study →
How authority is earned
Three tiers, set per task type, and you set the thresholds.
Every automated task runs at one of three levels of authority. The tier is set per task type rather than per system, so the same workflow can draft one step, wait for approval on another, and act alone on a third. A task moves up only when the evaluation suite shows the pass rate you defined as the threshold, and it moves back down the moment the monthly report shows a regression.
1
DraftedThe system prepares the work and a person finishes it. Every draft carries its evidence: the source documents, the rule or model that produced it, and its confidence. This is where every task starts, and where shadow mode runs before anything is shown to the team at all.Gated by the Exception Ledger — the task is not automated at all until its real deviations are written down.
2
ApprovedThe system prepares the complete action and a named person authorises it before anything consequential moves. The approval, the approver and the inputs go into the decision log.Gated by the evaluation suite — promotion from Drafted needs the per-step pass rate you set, measured against the golden dataset.
3
AutonomousThe system acts and the team audits afterwards, on a sample or on the exceptions. Any irreversible step stays at Approved regardless of accuracy, because a reversible mistake is a cost and an irreversible one is a liability.Gated by the monthly accuracy report — a regression alert sends the task back down a tier until the pass rate recovers.
Nothing is promoted on our say-so. The thresholds are yours, the report that tests them is a named line item, and the decision log shows every action that was taken at every tier.
The full method: five artifacts, three tiers, four rungs →How an engagement runs
Four rungs. You can stop after any of them.
0
Qualification call — free, 30 minutesA straight answer on whether there is anything here worth mapping. No deck, no proposal, no follow-up sequence.
1
Mapping Sprint — five days, with the people who do the workPaid discovery. Produces the Exception Ledger, the AI Boundary Map, the evaluation plan and either a fixed-price build quote or a written no. If it shows you nothing you did not already know, you do not pay.
2
Build — six weeks to productionScoped from what the sprint found. Majority deterministic software on top of your existing systems, every task starting at Drafted, shadow mode before anything acts.
3
Run — ongoing, and optionalMonthly accuracy report against the golden dataset, regression alerts, cost-per-run tracking. Stop the service and keep the system.
From the blog
The problems, one at a time.
Labor-Cost Congruity (DURC di Congruità) for Italian Builders 2026: How to Automate the CheckItalian-jurisdiction guide on automating the DURC di congruità (labor-cost congruity) check for construction firms. Italian only. Read at /it/blog/durc-congruita-manodopera-calcolo-automatico-edilizia/.
Construction Work Accounting & Measurement Books: How to Automate Them (2026)Italian-jurisdiction guide on automating construction work accounting (measurement book, register, SAL) under D.Lgs. 36/2023. Italian only. Read at /it/blog/contabilita-lavori-libretto-misure-automatica-edilizia/.
From Bill of Quantities to Automatic Quotes: Is It Possible in 2026?Complete guide to automating bill of quantities (computo metrico) from BIM models using PriMus, Edilius, AI-powered takeoff tools. How to generate quotes in hours instead of days.
Supplier Quote Comparison & RDO in Construction: How to Automate It (2026)Italian-jurisdiction guide on automating supplier quote comparison (RDO) in construction — request, offer normalization and comparison linked to the bill of quantities. Italian only. Read at /it/blog/comparativa-preventivi-fornitori-rdo-edilizia/.
Monthly Cassa Edile (MUT) Declaration for Italian Builders 2026: How to Automate It from PayrollItalian-jurisdiction guide on automating the monthly Cassa Edile (MUT) declaration from payroll data. Italian only. Read at /it/blog/denuncia-mensile-cassa-edile-mut-automatica/.
From RFQ to Subcontractor Contract: Automating the Tender CycleAutomated RFQ distribution, offer comparison, digital contract generation with qualified signature, D.Lgs. 36/2023 compliance, subcontractor qualification tracking.
Managing Delivery Notes (DDT) at Construction Sites: Automating Materials ReceiptDigital DDT capture with OCR, three-way matching (order-DDT-invoice), automatic reconciliation, supplier dispute resolution, warehouse integration, and invoice payment automation.
Construction Worker Time Tracking: From Clock to Automatic PayrollBiometric time clocks, geofenced mobile apps, BLE beacons, and automatic integration with Italian payroll systems (Zucchetti, TeamSystem) respecting CCNL Edilizia rules.
Managing Non-Conformances at Construction Sites: From Excel to Digital SystemsRNC (Non-Conformance Reports) with root cause analysis, corrective actions, photographic documentation, and automatic KPI generation for ISO 9001 compliance.
Public Tenders & Bids for Italian Builders 2026: Automating Bid Scouting & Offer AssemblyItalian-jurisdiction guide on automating public-tender scouting and offer preparation for construction firms. Italian only. Read at /it/blog/gare-appalto-bandi-pubblici-edilizia-ricerca-offerta-automatica/.
FAQ
Frequently asked questions
Which construction workflow do you usually start with?The one that costs the office an evening a week and has a document trail already: estimating from the price list, the DURC di congruità, or the SAL. The Mapping Sprint decides, because the documented process is never the real one and the exception that actually eats the time is usually not the one the owner names on the call.
Do we have to change our estimating or accounting software?No. We build on top of what you run, through its exports and APIs, and we put that in writing at the start. A replatforming is the one project in construction guaranteed not to pay back before the next season.
Who is responsible if a DURC di congruità calculation is wrong?A person, as today. The role drafts the calculation and shows its inputs; the office signs before submission. Every calculation, its inputs and its approver go into the decision log, so the answer to an inspector is a record rather than a memory.
How much of this is actually AI?Usually a minority. Of the six queues on this page, four are rules and lookups. The model is used where a document has to be read, drawings and supplier PDFs, and a rule does the pricing. Every step kept out of the model is cheaper to run and easier to defend.
Can a DURC or a Cassa Edile filing ever be sent without a person?Not from us. Anything that goes to an authority is an irreversible step and stays at the Approved tier regardless of accuracy: the role prepares the complete filing, a named person authorises it. Autonomous operation is reserved for reversible, internal steps and only after the evaluation suite shows the pass rate you set.
Is any of this fundable?Often. Process analysis and software development are recurring eligible costs in regional and national incentive schemes, and we work with a technical partner for finanza agevolata. Ask on the call and we will say what applies to your region and size.
Other sectorsAll sectors →
The same method, other queues.
ManufacturingThe order arrives forty different ways. The line needs it one way.
Wholesale & B2B distributionEvery order is a promise. Most of the work is checking whether you can keep it.
Thirty minutes to find out whether there is anything worth mapping.
The call is free, and we will tell you if the answer is no. Bring one queue that costs someone an evening a week.