Intent and context
The decision, workflow, users, information, consequences, and conditions the evidence represents.
AI System Evaluation
Can you prove this AI system is ready for real work?
A demo shows what can happen once. An evaluation shows what the system does under defined conditions, where the evidence stops, and who remains responsible when it is wrong.Standalone, fixed-scope work. Written evidence, not a score and a vibe.
The report
Five parts. Enough context to use the result.
The structure follows CAISI’s published reporting guidance so a reader can understand what was tested and where the limits are.
Read the published guidanceMethodology
Test cases, measures, protocols, and where people audited automated scoring.
System specifications
Model and version, settings, prompts, tools, connections, permissions, and approval points.
Interpretation
Supported conclusions, uncertainty, benchmark limits, failure patterns, and conditions.
Evaluator relationship
Who designed, implemented, paid for, and evaluated the system—including Lemonbrand.
Manager readiness
Operational commitments, not a compliance megaproject.
Most clients manage AI inside their business; they do not train foundation models. We assess the manager-side commitments in Canada’s Voluntary Code as a practical operating baseline.
AccountabilityNamed business and system owners.
Foreseeable impactsPeople, decisions, information, and misuse cases.
Oversight and feedbackMonitoring, feedback, escalation, and a stop route.
DisclosureClear identification where AI could be mistaken for a person.
CybersecurityProportionate review across access, data, integrations, and operation.
What you receive
A decision record your team can use.
Proceed, run a bounded pilot, repair and retest, obtain independent assurance, or stop.
AI System Evaluation Report
System and configuration record
Test protocol and evidence register
Failures, uncertainty, and limitations
Human-review and authority findings
Manager-readiness checklist
Conflict-of-interest disclosure
The boundary
Specific and true. Never a government endorsement.
CAISI does not certify Lemonbrand, this service, or the systems we evaluate. Following published guidance is not accreditation, approved-vendor status, legal advice, or independent assurance.
Put evidence behind the decision.
Bring the system, the decision it needs to support, and the consequence of getting it wrong.
YouTube
LinkedIn
TikTok
Instagram
Facebook
Substack