The Intelligence Layer for Construction
Addressing the AI technology gap construction is missing.
Vecreal is a vertical AI intelligence company for Canadian GCs. Project Intelligence connects the entire project and helps the PM deliver the project on-time and on-budget. Firm Intelligence carries that across all of a firm’s projects, so its hard-won knowledge compounds over time.
Software turned running the job into administering it
On a typical $20M project, a project manager tracks 200 to 300 live items at once: change orders, submittals, RFIs, subcontractor negotiations. Over thirty months these grow into hundreds of changes and RFIs and thousands of documents, spread across emails, schedules, drawings, and project records.
The load on a single mid-size project, the kind one PM runs. Figures describe a typical project of this size, not a named job.
The software currently used to manage all this is effectively a filing drawer. A project manager spends the day administering the record instead of running the work: filing documents, updating logs, drafting minutes, answering email, chasing responses. The judgement a PM is there for: the sequencing, the foresight, and the calls that protect the margin is in competition with admin work.
Construction has far more AI capability than it has put to use
Software that could read a thousand-page spec, a thirty-month email archive, a live schedule, and a change ledger, and reason across all of them to deliver each answer did not exist until recently. Frontier AI models can now do exactly that, across an entire project, at production quality. That is the technology Vecreal is built on.
The capability arrived faster than implementation, and construction is among the slowest to adopt it. AI can deliver tangible productivity gains, where construction has barely improved for decades. The capability-to-adoption gap is widest in exactly the work that runs a project:
Theoretical AI capability against observed use, percent, for the categories that fill a project manager’s day.
Source: Anthropic, Labor market impacts of AI, March 2026.
Construction looks like the wrong place to expect this. As a category it sits at the bottom of AI’s reach, because the trades are physical work no model can do. But a project manager works largely in the domain AI excels at. A PM’s day is management, finance, contracts, and engineering coordination, the categories at the very top of AI’s potential.
The reason that capability has not turned into results is not the model. It is context. A language model works by inference: asked about something it was never given, it predicts the most likely answer, and on the particulars of one project, where it has seen nothing, that confident prediction is exactly what a hallucination is.
Give that same model the project’s full record, its contract and drawings, its logs and email, and the guessing stops; it can reason directly from the record. Context is what turns a capable model into a reliable one in construction. Turning that into project performance takes three moves that build on each other:
- Memory: transform the project record into a database, so the scattered emails, letters, drawings, logs, and specs become one structured, queryable source that holds everything.
- Context: ground every answer in the live project record, each claim traceable to the document it came from.
- Workflow: put the grounded intelligence into the work itself, answering, triaging, drafting, flagging, rather than leaving it in a chat window on the side.
Do all three together, and capability finally becomes performance. For the PM, that reflects in a greater shift to high value work:
With the record-heavy load carried by Vecreal, more of a PM’s week goes to judgement, foresight, and client relationships, the highest-value work a GC has. That is how a firm improves both project and PM performance: by putting its best people where it matters most.
Project Intelligence reads the whole job and works from it
Our first product, Project Intelligence, reads across every tool the firm already uses and connects it all together. Where a generic model has to guess, Vecreal works from the actual project record. Each day, it:
- Prepares a morning brief on what changed overnight and what needs attention first.
- Holds the live state of the job, every open item in one place so nothing slips.
- Triages the email inbox, surfacing the most important correspondence and helping to clear it.
- Answers any question, including related correspondences, giving the complete picture.
- Drafts RFIs, submittals, and correspondence for the PM to send, with every claim cited to its source.
Project Intelligence does all of this without touching your systems. It prepares and drafts the work, but the PM stays in control of what gets sent. Every claim carries its source, click a citation and the record opens, and if a claim can’t be traced, it doesn’t ship. The models retain nothing and train on nothing, and the data stays in Canada. We implemented careful boundaries because we understand the risks of being wrong in construction.
The Workshop is built for any deep work. The PM can hand Vecreal any task, an RFI to answer, a change order to price, a contract clause to check, and Vecreal works it against the contract and the record and produces drafted responses. Vecreal is built to work with the PM on long-running, difficult tasks.
Together, these give the PM back the job: the day’s priorities, the live state, the drafts, and the answers, every one drawn from the project’s own record and traceable to it. The administrative load that crowded out the work is better handled, and the judgement a PM is there for gets priority.
Learn more about Project Intelligence →
Every project compounds the firm’s intelligence
What Project Intelligence holds for one job becomes worth far more across many.
Every project’s decisions, lessons, and live state flow up into one firm memory, and the firm layer reads across all of them at once.
Three kinds of intelligence surface that no single project can see:
- Cross-project patterns. The scope gaps and contract misses that quietly repeat across jobs, visible only when every project is read together. What do we keep underbidding?
- Counterparty memory. Every sub, vendor, and consultant evaluated as the work happens, with the trend in hand before the next award. Which of our subs are quietly getting worse?
- Institutional knowledge. The lessons a firm spends years paying to learn, held as permanent memory that outlasts the people who learned them. What did our best people know that we lost?
This is the real loss. When a senior PM leaves, the firm’s deepest expertise has always left with them, and roughly 263,400 Canadian construction workers are set to retire by 2033 (BuildForce Canada, 2024). Firm Intelligence is the layer that keeps it: a new project opens with the firm’s relevant lessons, live policy, and counterparty history already loaded, and a new hire inherits all of it instead of starting from a blank slate.
The firm’s edge over time: with Firm Intelligence, every project’s lessons are retained and compound; without it, memory stays low and drops with each senior departure.
Firm Intelligence can only emerge from Project Intelligence running across your projects, turning high-variance, unstructured data into synthesized learnings. The project layer earns its keep in the first month. The firm layer is why we built the company.
Learn more about Firm Intelligence →
Improving core outcomes for GCs
Project Intelligence and Firm Intelligence are the solution. They improve core outcomes for GCs on projects: protected margin, schedules held, stronger contract positions, and more capacity in every PM. It works by introducing intelligence through the actual execution of the work.
It doesn’t arrive as another dashboard or one-off tool. It shows up in daily briefs, deep-work sessions, and in structuring complex project data. And it compounds: what Project Intelligence produces on each job feeds Firm Intelligence, which carries the gains across the firm.
I built this because I have run both sides of it
I spent years managing construction projects across Canada: labs, fit-outs, hospitals, schools, high-rises, complex renovations, mission-critical infrastructure, for general contractors and as an owner-side PM. I have also built and led a software company in financial infrastructure, and today I build my own agentic AI systems that run parts of my work and my company. Vecreal is where those two careers meet: built for the people who carry the job, and for firms to compound their performance over time.
Having built and implemented these systems myself, I understand how they fail and where they can be maximally leveraged. Vecreal is built to operate to the high standards GCs require, not as a quick patch solution.
The capability is here, and construction has barely begun to use it. The firms that close that gap first compound an advantage that is hard to buy back: a memory that deepens with every project, pulling the firm further ahead the longer it runs.
Vecreal is the intelligence layer construction has been missing. It’s built on how actual GC projects perform, and to the high standards required. We are currently working with a small group of Canadian GCs on pilot projects. If your firm runs complex work and is ready to implement AI, let’s connect.
