Grounded AI assistant
Medi
A grounded place to ask what comes next.
Medi works inside Monitor with approved knowledge and permitted workflow context. It explains requirements, shows the basis for an answer, and escalates when policy, uncertainty, or judgement calls for a person.
Medi supports compliance workflows. It is not a medical-advice service and does not make autonomous final compliance decisions.
Grounded in the work already happening.
Medi sits beside the Monitor record. It helps people understand requirements and status while keeping consequential decisions in the governed workflow.
Ask in plain language
Students, staff, and administrators can ask about requirements, status, and the next step without learning a new product vocabulary.
Retrieve approved context
Medi uses the relevant knowledge base and permitted Monitor context instead of roaming the open web for an answer.
Explain the basis
The response is grounded in approved content and designed to show the source or record context behind the explanation.
Escalate when people should decide
Uncertain, consequential, exceptional, or customer-defined questions move to an accountable human path.
Research & evidence
Models change. Medi’s standard stays fixed.
Medi Bench is the canonical research record for comparing grounded quality, cost, and response speed before a model is considered for governed workflows.
Archived May 2026 evidence
Full methodologyV1 reported grounded-answer rate, cost per accepted answer, and p95 latency. It is archived directional evidence with no published corpus size, repeats, or confidence intervals, and is not directly comparable with the planned V3 weighted methodology.
| Model | Grounded answer rate | Cost per accepted answer (USD) | p95 latency | Confidence | Run |
|---|---|---|---|---|---|
| 99% | $0.0040 | 9.7s | The archived aggregate fixture did not publish confidence intervals. | Not published · Not published prompts · No repeats published | |
| 100% | $0.0008 | 8.1s | The archived aggregate fixture did not publish confidence intervals. | Not published · Not published prompts · No repeats published | |
| 98% | $0.0035 | 10.6s | The archived aggregate fixture did not publish confidence intervals. | Not published · Not published prompts · No repeats published | |
| 95% | $0.0005 | 14.3s | The archived aggregate fixture did not publish confidence intervals. | Not published · Not published prompts · No repeats published |
- Displayed
- V1 · archived
- Latest candidate
- V3 · pilot
- Artifact
- Signature verified
PILOT / NOT FOR PUBLIC CLAIMS. V3 model results are pending and not published. The chart uses only the four rounded values displayed in the May 2026 V1 chart; it includes no later models, invented secondary metrics, or confidence intervals.
Why the name
A small name with a personal beginning.
The name came from a weekend idea between Compliance Health Founding CEO Matthew Protti and his daughter. Medi is short for Medical or Medicine, but it is also intentionally simple, friendly, generic, and easy to pronounce.
- A name people can say without a product manual
- A character that makes the AI assistant visible
- A friendly guide for work that can otherwise feel heavy
Why a polar bear
A Canadian character with a Bow Valley connection.
Bow Valley College was part of the inspiration from the beginning. Its esports identity uses a bear, and the polar bear gave Medi a distinctly Canadian, recognizable form while making one thing obvious: you are talking with an AI assistant.
- Canadian by character and operating context
- Connected to the people and institutions that shaped the product
- Designed to feel approachable without pretending to be human
Helpful by design. Bounded on purpose.
The useful part of an AI assistant is not simply producing an answer. It is knowing which knowledge is allowed, what context is relevant, and when a person needs to take over.
Approved knowledge first.
Medi is designed to answer from controlled knowledge and permitted Monitor context, not unsupported recall.
Clearly an assistant.
Medi explains and supports. It does not make autonomous final compliance or medical decisions.
People stay accountable.
When confidence, policy, or judgement calls for a person, Medi makes the handoff explicit.
See how Medi fits your Monitor workflow.
Bring one real question path. We will show how approved knowledge, Monitor context, citations, and human escalation can work together.