Canada's Construction Intelligence Platform.
CCAIP turns building codes, drawings, BIM models, permit documents and municipal review workflows into a traceable, deterministic, human-reviewed digital process.
Bedroom egress window opening area not provided
NBC 2020 · Div. B · 9.9.10.1.(1) — Egress Windows or Doors for Bedrooms
Illustrative example. Not real system output.
The problem
Permit review is still a manual reading exercise.
A permit package arrives as drawings, models, forms and supporting documents. Somebody has to read all of it, hold the applicable code in their head, decide what applies, and write up what is missing. It is slow, it is inconsistent between reviewers, and the reasoning behind a decision is rarely recorded in a way anyone can re-examine later.
Review capacity is the bottleneck
Applications queue behind a finite number of qualified reviewers, and every resubmission re-enters the same queue.
Two reviewers, two answers
The same package can attract different comments depending on who picks it up and how much time they have.
The reasoning disappears
A decision is recorded, but the facts, provisions and evidence that produced it usually are not.
Module One
Automated Compliance Checking for building permit applications.
CCAIP is a platform, not a single tool. Automated Compliance Checking is the first module: ingest a permit package, normalize it into facts, run deterministic checks against machine-readable code provisions, and hand a reviewer a findings list where every result points back at the evidence that produced it.
What goes in
2D permit drawings (PDF), BIM and IFC models, structured and unstructured permit application data, and digitalized construction codes — including submissions delivered as a single package.
What comes out
A reviewer-ready findings list with a category, a code citation, an explanation, the facts evaluated, the facts missing, and the evidence chain — exportable as a report, a spreadsheet, structured JSON, or through the API.
How it works
Facts flow into rules. Rules produce auditable outcomes.
Nothing in the pipeline is a black box. Each stage produces an artifact the next stage consumes, and each artifact can be inspected on its own.
Intake
PermitFlow versions the submission and validates the package before anything is processed.
Extraction
Text, sheets and title blocks from drawings; storeys, spaces, doors, windows, stairs and properties from IFC.
Normalization
Extractions become typed building facts — heights, areas, counts, ratings — in a building knowledge graph.
Execution
CodeEngine runs machine-readable rules against those facts. Deterministic, versioned, repeatable.
Review
ReviewDesk gives officials the findings queue: confirm, comment, override, or request information.
Export
Reports, findings exports and structured results, through files or the REST API.
Four results. No others.
Every rule execution returns exactly one primary category. Missing information is not failure, and uncertainty is stated rather than smoothed over.
MEETSThe evaluated facts satisfy the provision.
DOES_NOT_MEETThe evaluated facts contradict the provision.
INFORMATION_NOT_AVAILABLEThe submission does not contain the facts the rule requires.
UNCERTAINThe facts are present but ambiguous, conflicting, or low confidence.
Compliance doctrine
AI assists. Rules decide. Humans approve.
This is the line that makes the platform defensible, and it is enforced in the architecture rather than in a policy document. A language model is useful for reading a drawing. It is not an authority on whether a building is compliant.
AI may
- Extract data from drawings, models and documents
- Classify buildings from extracted facts
- Search the code corpus
- Explain a result in plain language
- Estimate confidence and flag ambiguity
AI may never
- Approve compliance
- Reject compliance
- Invent a citation
- Silently override a deterministic result
- Replace the judgement of a building official
Compliance categories are produced only by CodeEngine. Missing information is not failure. Uncertainty is explicit. Human reviewers retain final authority.
Platform
Named components, not features.
Each component is separable, API-addressable, and multi-tenant from day one — so the same architecture can carry later modules without being rebuilt.
Six platform-wide commitments
These are architecture requirements across the whole platform, not per-module options.
Official languages
English and French code sources and bilingual operation designed in, not bolted on.
Sovereign routing
Fully local, fully hosted, hybrid, reverse-hybrid and restricted deployment profiles, including local model endpoints.
Verifiable audit
Append-only IntelicoreAI Ledger evidence, so an audit trail can be checked rather than trusted.
National federation
Multi-jurisdiction by design — no province, code edition or AHJ hard-wired into the core.
Compliance intelligence
Code and rule versioning, permit history profiles, scoring and trend analysis across submissions.
Governed AI
Provider-independent AI assistance with governance controls and a supported no-LLM mode.
Who it's for
Built for the people who carry the review.
Municipalities and AHJs
Triage submissions faster, comment consistently, and keep a defensible record of why.
- Completeness screening at intake
- Findings queue with overrides
- Official review report exports
- Audit trail on every action
Architects and designers
Run the check before you submit, not after the comments come back.
- Pre-submission compliance check
- Missing-information list
- Code provision lookup
- Fewer resubmission cycles
Builders and developers
Reduce the delay risk that sits between a design and a shovel in the ground.
- Permit readiness reporting
- Clear, itemized issue lists
- Faster correction turnaround
- More predictable timelines
Engineering firms
Discipline-specific checks with the evidence attached to each result.
- Fact-level traceability
- Reviewable rule logic
- Structured exports
Software vendors
API access to deterministic compliance and code intelligence, without rebuilding either.
- REST API access
- Structured findings and evidence
- Sandbox environment
Provinces and territories
Digital code management and amendment tracking across jurisdictions.
- Versioned rule packs
- Amendment change detection
- Multi-jurisdiction federation
Research context
Developed against the NRC automated compliance checking challenge.
CCAIP Module One is being developed against the National Research Council of Canada's Innovative Solutions Canada challenge for automated compliance checking of building permit applications. The prototype is designed to demonstrate a web-based system that can:
- Ingest 2D permit drawings and BIM/IFC models
- Ingest structured and unstructured permit data
- Ingest digitalized construction codes
- Normalize building facts into a knowledge graph
- Run deterministic compliance checks
- Use AI only for extraction, classification, explanation and uncertainty
- Produce meets, does-not-meet, missing-information and uncertain results
- Trace results to provisions and source evidence
- Support human-in-the-loop review
- Export results through files and APIs
- Support Canadian data residency
- Support English and French code sources
Contact
Talk to us about a pilot.
If you run a building department, design permit packages for a living, or build software that touches permitting, we would like to hear how your review process actually works.