Our read
Named contractors document three separate jobs: progress and delay forecasting (JE Dunn, NCC with Buildots), visual records, QA/QC and owner reports (NOVO, Suffolk with OpenSpace), and RFIs, submittals and document questions (Suffolk, Torcon with Trunk Tools). Pick the job first. Sources do not show these tools beating each other, and a logo alone shows adoption, not use.
Which tool for which job?
| Your job | Documented example | The reference to read | What it documents |
|---|---|---|---|
| Forecast delays and move manpower |  Buildots |  JE Dunn | JE Dunn compares actual progress with the schedule in Delay Forecast to spot bottlenecks and reallocate manpower, and settles AI flags in the Explorer tool. 1 |
| Hold trades to weekly plans | Buildots |  NCC | NCC compares each subcontractor's completion rate and pace with weekly plans and analyses pace risks on a bi-weekly basis on Finnish projects. 5 |
| Draft weekly owner reports |  OpenSpace |  NOVO Construction | NOVO Construction has an AI agent pick photos from pinned capture locations and draft its weekly owner report, which it says is 85-90% complete before review. 2 |
| Document site conditions on every job | OpenSpace | Suffolk | Suffolk expects OpenSpace capture at least every two weeks and uses it for coordination meetings, owner updates, punch and preconstruction. 6 |
| Answer field questions and draft RFIs |  Trunk Tools | Suffolk | A Suffolk superintendent asks Trunk Tools for answers on a phone and has it draft RFIs, which the superintendent vets before sending. 3 |
| Review submittals against specs | Trunk Tools |  Torcon | Torcon ran TrunkSubmittal on 64 reviews across two projects; most came back non-compliant (69%) or partially compliant (30%). 7 |
Which jobsite jobs do named contractors actually use AI for?
Three jobs recur in named contractor sources. JE Dunn and NCC use Buildots to track progress against the plan and forecast delays. NOVO Construction and Suffolk use OpenSpace as a visual jobsite record behind QA/QC, owner communication and reporting. Suffolk and Torcon use Trunk Tools to answer questions from project documents, draft RFIs and review submittals. These are examples to investigate, not a ranking of tools.
As of 2026-11-02, usedby lists 37 construction companies with a tool in the industrial and construction AI category: 11 CONFIRMED by a case study, quote or press release and 26 OBSERVED on a vendor's own pages. Those counts show who is named and at what level. They say nothing about which tool performs better.
Augury and Neural Concept sit in the same category but do different work: Augury targets machine health in manufacturing plants, and Neural Concept targets CAD and simulation for OEM engineering teams. Holcim (Augury) and RWDI (Neural Concept) are shown by those vendors as customers, and no task is documented for either.
Progress tracking and delay forecasts: what do JE Dunn and NCC show?
JE Dunn used Buildots on a high-rise residential project that began as a quality-control measure. According to the Buildots case study, it grew into forecasting, communication with trade partners and automated day-to-day supervision. Project managers used the Delay Forecast to compare actual progress with the schedule and reallocate manpower before a delay occurred. The team reconciles AI-raised flags in the Explorer tool instead of running manual photo walkthroughs. The study reports 3 million square feet tracked across JE Dunn projects.
NCC applied Buildots on six Finnish projects covering 172,000 square metres. At a school renovation, the platform flagged installations that did not match the model, so the team corrected the work or updated the model. On other projects, managers compared each subcontractor's completion rate and pace with weekly plans and analysed pace risks on a bi-weekly basis, as the source words it. The source says weekly Percent Plan Complete improved over time, without a figure.
What transfers is a routine, not a feature. In both stories, site captures are compared with a plan, and a manager then acts on the result: reallocating labour, correcting an installation, or holding a trade to a weekly commitment. Editorial recommendation: before a pilot, decide who reviews the flags each week and what happens when a flag disagrees with the superintendent.
Visual records, QA/QC and owner reports: what do NOVO and Suffolk show?
NOVO Construction has used OpenSpace since 2017 and runs it on every substantial project, more than 100 active jobs. Teams use it for QA/QC, owner communication, warranty documentation and routine site walks. NOVO built an AI agent on top of that record for the weekly owner report, a 10-to-15-page update that used to take most of a workday. Regular capture locations became points of interest pinned to drawings, and the agent selects photos, recreates viewpoints and drafts copy in NOVO's format. NOVO reports the draft is 85 to 90 percent complete before review.
Suffolk's OpenSpace case study describes a different scale. OpenSpace is expected on every job, with capture at least every two weeks, and Suffolk uses it in coordination meetings, owner communication, punch and closeout, and preconstruction existing-conditions documentation. Drone imagery, field notes and progress tracking are added in stages. Progress tracking was in early implementation on a live project, with the rollout described as disciplined rather than fast.
The lesson from both is sequence. NOVO's agent works because years of tagged, located captures already existed. If your teams do not yet capture on a fixed cadence, that habit comes before any agent. OpenSpace lists integrations with Procore, Autodesk, Revizto and schedule imports from Oracle P6, Asta Powerproject and Microsoft Project, so check those against your own stack.
Can AI take over RFIs, submittals and document searches?
A Suffolk superintendent, Andrew Bared, uses Trunk Tools on a phone for answers to subcontractor questions, to check RFIs against ongoing work, to see where inspections are missing, and to draft RFIs. In the interview, the superintendent vets each draft before it goes out. Torcon piloted the tool from August 2025 on two projects with TrunkText for source-backed questions and TrunkSubmittal for compliance reviews. The vendor-published figures are 453 hours saved over four months and 64 submittal reviews, most flagged non-compliant (69%) or partially compliant (30%).
Gilbane Building Company's record is a quote from a senior superintendent, Andrew Roy, who describes TrunkText doing the document digging for them so they can be more responsive in the field. That source gives no hours or task counts, so it cannot be set beside Torcon's figures.
One restriction applies to all of them: the tool drafts, and a person decides. Trunk Tools lists SOC 2 Type II among its compliance items, and its trust page names subprocessors; read both against your data policy before uploading drawings and specs.
Which named peers are in the US and Canadian record, and at what level?
CONFIRMED with a documented line: JE Dunn (Buildots, forecasting), NOVO Construction (OpenSpace, QA/QC and owner reports) and Suffolk (OpenSpace and Trunk Tools). Gilbane Building Company (Trunk Tools) is CONFIRMED by a superintendent's quote without a described task. OBSERVED, with the vendor showing the company and no task recorded here: Juneau Construction (Buildots) and RWDI in Canada (Neural Concept). A logo establishes adoption, not a workflow.
What should a smaller contractor check before copying one of these workflows?
The stories involve large builders: NCC cites 14,500 employees and Suffolk 3,500, and Suffolk and NOVO describe close work with the vendor's onboarding staff. Whether a smaller firm gets the same support, effort or results is not shown. Test three things: a capture routine your crews will keep, a named owner for the output, and an integration with the project system you already pay for. Pricing is unknown in the sources reviewed.
Method and limits
2026-11-02
Scope: 37 companies classified in Construction on usedby with a public line on a Industrial & construction AI tool. Every count on this page covers this sector only, not the tools' whole customer base. Company size is not recorded and is not claimed.
This dated snapshot counts distinct companies in public customer evidence. CONFIRMED means a case study, named testimonial or press source; DECLARED means the company itself told usedby; OBSERVED means a vendor listing; SIGNAL means a job post. All four count, but their strength differs. Removed and unlevelled lines are excluded. Examples are a capped selection verified within 90 days. Vendor facts are sourced records checked within 60 days. Coverage is uneven and missing evidence is unknown. These figures measure neither market share nor product performance; capture dates are not adoption dates.
