Home Compliance Responsible AI
AI that informs. People who decide.
How every AI system we deploy is scoped, grounded, tested and overseen. Written so an evaluator can score it line by line, including against responsible AI criteria in federal and municipal solicitations.
Seven commitments, in every deployment.
These are defaults, not options. They appear in our statements of work, and each one answers the objection evaluators raise most: what stops the system from making things up, and who is accountable when it matters.
- Scope by default
-
Deployments are informational-only. No automated administrative decisions: the system never approves, denies, prioritizes or adjudicates anything. A human decides, always.
- Grounding
-
Retrieval-only over curated, client-approved sources. Every answer cites its source, so staff and residents can verify it. The system declines questions outside its corpus rather than guessing.
- Human in the loop
-
Every deployment includes an escalation path to a named human. When the system cannot answer, or the visitor asks for a person, it hands off rather than improvising.
- Testing
-
Bias and accuracy testing before launch, then monthly in production, with results reported to the client. Accuracy is measured against the approved corpus, not against impressions.
- Residency
-
Models and data are configured for Canadian residency per engagement, named in the statement of work.
- Ownership
-
Clients own the prompts, configurations and outputs of their deployment. There is no lock-in by design.
- Alignment
-
Our deployments are designed to align with the federal Directive on Automated Decision-Making and the federal FASTER guiding principles for generative AI. Because our systems make no administrative decisions, they sit deliberately below the Directive's risk thresholds.
Statement prepared August 2026
Where these commitments ship.
This statement is not aspirational. It describes two priced offers you can buy today.
Bilingual AI Answer Bot
From $25KA retrieval-only chatbot over your own published content, English and French by default, with every commitment on this page built in.
How the answer bot worksAI Proof of Concept
$39.5K fixedAn 8-week AI use case implementation with success criteria defined before build, scoped informational-only from day one.
How the proof of concept worksEvaluating an AI purchase?
We will walk through this statement against your evaluation criteria, line by line. RFI answers are free.