AI agent platform · Construction · Québec
AI agents do the work.
You keep the GO.
Ino-Vita AI runs the workflows between your email, your accounting, your drawings and your job sites. Every action cites its sources and is checked by independent AI models — and nothing external goes out without human sign-off. Built by a Montreal general contractor who runs it on his own operations, every day.
Every demo is a live system — not a slide deck. · 514 922-3336
The problem
The truth lives in five systems — and in nobody's head.
For a contractor with 5 to 100 employees, the real cost isn't the software. It's the coordination between the software.
Email, drawings, job site, accounting: nothing talks to anything
The context behind every decision is scattered across the inbox, the drawings, the field tools and the books. Nobody has the full picture at the moment it matters.
Your managers spend the day chasing and retyping
Collecting, following up, double-checking, re-keying information from one system to the next: low-value work with a high cost of error, eating the management capacity you actually pay for.
A generic AI assistant accelerates the mistakes too
Without cited sources, validation or sign-off, generic AI can speed up errors as fast as it speeds up work. In construction, a wrong amount or a missed directive is real money and real liability.
How it works
A governed orchestration layer, on top of the tools you already use.
We don't replace your software. Specialized AI agents run the full loop — and stop in front of a human for anything external or financial.
Control principles — built in, not bolted on
- Read first: no action without cited sources
- Independent cross-validation — the author never grades its own work
- Human approval gates on every external or financial action
- Full audit trail and an emergency kill switch
- Client data isolated, client by client
The human layer — construction people
- Sales and diagnostics by people who know construction
- Implementation and configuration on your actual process
- Training, adoption follow-up, one named person you can reach
- Escalation to a human, always
Try the gate
Take the boss's seat. Nothing leaves without your GO.
A 60-second simulation: a fictional email comes in, the agents do their work — then stop in front of you. Your call.
"Hi, further to the requested changes, please find our revised quote at $9,800. Kindly confirm so we can start production."
Attachment: revised-quote.pdf (fictional)
3 sources linked: the email thread, purchase order PO-0000 (fictional, $9,200) and drawing rev.04. Discrepancy flagged: +$600 vs the purchase order.
This draft would go to a supplier: that's an external action. The machine stops here. Your move.
That's governance: the AI does the work, you keep the decision.
Want to see the real machine, on our real operations?
Book a 30-minute callProof
Not a prototype. Our own company runs on it.
The platform was built to operate a real general contractor — ours. Here's what it does for us, backed by dated internal audit figures.
Our rule on numbers: no client-performance claim ("time saved", "errors reduced") gets published before it's been measured by our formal program — metric, period, scope and primary source defined. The figures above are internal operating facts, audited and dated.
Rollout
Three stages, each with its own exit door.
No big bang: every stage has measured criteria before moving to the next — and a gradual go-live, on your actual process.
We start from your actual process
- Diagnostic of how you really work; pick the first high-value workflow
- Pilot scope agreed and signed
- Baseline captured — the point of comparison
Gradual go-live, with you
- Connection to your systems; rules, gates and approvals configured
- Testing, training, progressive go-live
- Criteria to advance: adoption at the agreed threshold, error rate within tolerance
The governed service, day to day
- Ongoing operation, monitoring, updates and support
- Named ambassador: adoption follow-up, escalations, quarterly value review
- Additional workflow domains activated at your pace
The exact timeline depends on your first workflow and your systems — it's set at the diagnostic, not promised up front.
Plans
One implementation. One governed service. One named person.
Every client starts with an implementation on their actual process, then moves to the recurring service: the platform operates, the human team stays reachable.
- 1 domain (e.g. email + documents)
- Monitoring and support
- Quarterly review
- 3 domains (e.g. + accounts payable, site tracking)
- Named ambassador
- Monitoring, support, quarterly review
- All domains
- Dedicated ambassador hours
- Custom gates and reporting
Exact investment and scope are confirmed at the diagnostic, based on your first workflow and your systems. Prices in Canadian dollars, before taxes.
Straight questions
The questions we actually get asked.
Q1Where does my data go? Who sees it?
Each client gets an isolated deployment: your data is never mixed with another client's. Agents work "read first" — they cite their sources —, every action lands in the audit trail, and an emergency kill switch cuts everything in one move.
These are the same rules we impose on our own construction operations, every day.
Q2Does this replace my current software?
No. The platform sits on top of your existing tools and does the work between them. At our company it orchestrates email, accounting, project management and document storage (Gmail, QuickBooks Online, Procore, Google Drive). Your tools remain the source of record; the agents do the coordination.
Q3What if the AI gets it wrong?
That's the question the system was designed around. Three locks, in order: (1) every significant deliverable goes through cross-validation by independent AI models, from different families than the author — the author never grades its own work; (2) nothing external or financial goes out without human approval; (3) everything is logged, so it's verifiable after the fact.
This mechanism has already caught real, documented errors on our side — before they reached a third party.
Q4Where does my team fit in?
At the center. The agents handle the collecting, preparing and following up; your people keep the decisions — and every client has one named, reachable person on our side: diagnostics by construction people, training, adoption follow-up, escalation to a human, always.
Q5How long before it's running at my company?
The rollout is gradual: diagnostic and baseline, then a guided pilot on a first workflow, then production. The timeline depends on your first workflow and your systems — it's set at the diagnostic. We don't promise a standard timeline before we know your process.
Q6Why should we believe you?
Because we're not asking you to believe us: the demo is our own system running on our own construction operations. And we won't publish any client-performance figure before it's been formally measured — metric, period, scope and source defined. What we show today are our internal facts, audited and dated.
About
Born on real job sites, not in a lab.
Groupe GIV / Ino-Vita is a Montreal construction group delivering institutional and commercial projects — schools, public buildings, multi-residential. In 2025–2026, the group hit the same wall as every mid-size contractor: coordination was devouring management capacity.
Instead of buying yet another piece of software, the team built a layer of specialized AI agents on top of its existing systems — governed by explicit rules: read first, sources cited, independent validation, human approval on anything external or financial, full logging, emergency stop.
That platform operates the company every day. The venture exists to industrialize it and offer it to contractors with 5 to 100 employees: AI agents that do the production work, surrounded by a small human team that sells, configures, trains — and stays reachable.
§09 · Next step
Book a 30-minute intro call.
A real construction person, not a form. We look at your first high-value workflow together — and show you the system running on our own operations. No commitment.
Email: sarah@ino-vita.com Montréal, Québec