Navigate

No Integration Required to Start

Deployment in days, not months - without touching existing systems.

COMPAiSS does not require integration with internal systems, databases, or CRMs to operate. It uses publicly available institutional sources combined with a governed authorization layer to deliver accurate, policy-aligned responses from day one.

There is no data ingestion pipeline to build. No access to internal records is required. No IT project resourcing or system changes are needed before a deployment can begin.

Deployment can be completed in days. An institution can evaluate COMPAiSS against its own live questions, using its own publicly available content, without any dependency on internal systems or approvals from technology teams.

This distinguishes two phases of deployment. Initial evaluation and pilot access, where an institution begins testing COMPAiSS against its own questions, can be completed in days. Full production deployment, including greenlist scoping, staff pilot review, and institutional sign-off, is typically completed within 2-3 weeks of contract execution (see Higher Education Institutional Pricing below).

The system is ready to operate the moment institutional sources are authorized. Nothing else needs to be connected first.

Compatible with Existing CRM Workflows

Integration at the interaction layer - not the data layer.

For institutions that want to extend usage into operational workflows, COMPAiSS can operate alongside existing CRM platforms - including student information and admissions systems - without accessing or modifying internal data.

This distinction matters. Most AI vendors approach CRM integration at the data layer - connecting to records, ingesting structured content, and embedding inference directly into the system architecture. That approach introduces dependency, risk, and ongoing maintenance obligations.

COMPAiSS integrates at the interaction layer. Staff continue working in their existing systems. When complex or policy-based questions arise, COMPAiSS generates accurate, policy-aligned, ready-to-use responses that can be returned directly into staff workflows.

Standard CRM AI integration
  • Embeds AI directly into the CRM interface
  • Requires access to student records and internal data
  • Generates answers from fragmented or constantly changing content
  • Inference-first: generates first, validates afterward
  • Introduces architecture dependency and ongoing maintenance
COMPAiSS alongside your CRM
  • Operates alongside the CRM - not inside it
  • No student records or internal databases required
  • Generates answers only from institution-authorized sources
  • Authorization-first: governs before generating
  • No system dependency - can be adopted or discontinued independently
COMPAiSS governs how information is used - not how data is stored. It enhances existing systems without becoming part of the system architecture.

This also means that tools like Copilot or other productivity AI remain useful for drafting and communications. COMPAiSS addresses a different problem: ensuring that the content of those communications is policy-correct and institutionally authorized.

Copilot helps you write faster. COMPAiSS helps ensure what you write is correct.

No Enterprise Integration Burden

The cost and complexity that typically accompany enterprise AI adoption are eliminated by architecture.

Large-scale RAG deployments and Copilot-style enterprise AI systems carry a common set of infrastructure requirements that drive cost and implementation timelines. COMPAiSS does not.

Not Required

Data ingestion or synchronization pipelines. Vector databases or indexing infrastructure. Access to internal systems or APIs. Ongoing content cleaning or maintenance cycles.

What This Means

No IT project to initiate. No vendor lock-in. No infrastructure investment before evaluation. No dependency on large data engineering teams or CRM restructuring.

The Result

Lower implementation risk. Faster path to evaluation. Cost structure that scales without requiring major infrastructure investment. A system that can be discontinued at any time without consequence.

This is not a trade-off against capability. It is a consequence of a different architectural approach - one in which the authorization layer, rather than the data layer, does the work of ensuring accuracy and scope compliance.


Greenlist Control, Dashboard Access, and Ongoing Source Auditing

Institutions retain full control over every source COMPAiSS is authorized to draw from - with transparent tooling and ongoing maintenance built in.

The greenlist is the governance foundation of every COMPAiSS deployment. It defines precisely which institutional sources the system is authorized to draw from - and nothing else. Institutions retain full visibility and control over this list throughout the deployment.

📋

Greenlist Dashboard

Authorized institutional staff access a secure, browser-based dashboard to view, add, and manage approved sources directly. No technical expertise required. Changes take effect within seconds and are immediately reflected in system responses. The complete version history of every modification is available for audit. For the technical mechanics behind this, including redirect classification, canonical URL normalization, and the automated deployment pipeline, see Greenlist Integrity and Maintenance on the AI Governance page.

🔍

Ongoing Source Auditing

The full greenlist for each institution is audited regularly - checking for broken links, outdated pages, redirects, and sources requiring update or replacement. Audit results are shared with the institution directly.

No IT Dependency

Source gaps identified during testing are closed within hours. Broken or outdated links are caught before they affect student responses. No dependency on vendor timelines or internal IT queues to update institutional content.

What this means operationally: The institution controls exactly what the system knows and cites. During pilot deployments, a single authorized contact manages the list in coordination with COMPAiSS to ensure consistency and prevent errors. As the deployment matures, greenlist management transitions naturally to the institution's own designated staff.

This is not a background technical process - it is a visible, auditable governance control that procurement officers, IT reviewers, and senior administrators can inspect, verify, and rely on.

Why Greenlists Scale Better Than They First Appear

The most common objection to greenlist governance is that it will not hold up at scale. That objection gets the scaling problem backwards.

The question is not how many URLs a system can store. Modern software manages millions of records without difficulty. The real question is whether institutional ownership scales, and it already does, because every large institution has already solved this problem for everything except AI.

Universities distribute responsibility for admissions content to the Registrar, financial policy to Financial Aid, and academic requirements to each faculty. Hospitals distribute patient education content to Clinical Governance, medication information to Pharmacy, and infection control guidance to Infection Prevention. None of that ownership was created for COMPAiSS. It already exists, department by department, and has existed for years.

The greenlist does not build a new governance structure. It mirrors the one that already runs the institution. Every institution already knows who owns each policy. COMPAiSS simply asks the AI to respect those existing ownership decisions. As a deployment grows, from one faculty to an entire university, from one hospital to a regional health network, governance scales the same way it always has: by adding the next already-existing content owner to a list, not by re-engineering the system underneath them.

This points to the underlying trade COMPAiSS makes, and it is worth stating plainly. Conventional retrieval systems are built for ingestion convenience: point a crawler or connector at a document estate and let it index everything reachable. That convenience comes at a cost, nobody can say with confidence which of those documents were actually reviewed and approved for AI use, only that they were technically reachable. COMPAiSS makes the opposite trade. It gives up ingestion convenience in exchange for provenance control: knowing, for every source the AI can cite, exactly who authorized it and when. Greenlists do not create new governance work. They make existing governance work explicit, attributable, and auditable.

That distinction also changes what a governance failure looks like when it happens. In a conventional retrieval system, a governance failure, an outdated policy, a broken link, a document nobody meant to include, typically surfaces as an incorrect answer delivered to a user, discovered only after the fact. In a COMPAiSS deployment, the same underlying failure surfaces differently: as a missing answer flagged for correction, caught by the validation cycle before it ever reaches a user.

Conventional Retrieval COMPAiSS
Failure Mode An incorrect answer, delivered to a user A missing answer, flagged for correction before delivery
Detection After a user acts on it During routine validation, before any user sees it
Correction An incident, someone notices the answer was wrong A dashboard alert, before the content is ever cited

This is the same failure, addressed at two different points. One architecture discovers it downstream, after harm. The other discovers it upstream, before generation.

The greenlist is not merely a configuration file. It is the institution's documented authorization boundary for AI.
Hard Questions About Greenlist Governance

Direct answers to the objections procurement officers, IT security teams, and governance committees raise most often.

Does this scale past a few thousand sources?

Scale comes from delegation, not from bigger software. A provincial health authority with sources spread across twelve hospitals does not need twelve times the governance effort of one hospital. It needs the same model, replicated across twelve clinical governance offices that already own that content today. Each site manages its own portion of the greenlist independently, using the same dashboard and audit cycle regardless of whether the list holds hundreds of entries or tens of thousands.

Who approves sources, and who audits those approvals?

Every addition, removal, and modification to the greenlist is logged with a timestamp and attributed to the person who made it. That is the entire point: governance decisions become attributable instead of invisible. Today, in most institutions, an AI-relevant content decision is effectively made by whoever configures the crawler or connector, with no department sign-off and no record of that decision at all. A greenlist error is visible in the change history and correctable within minutes. The equivalent error in conventional retrieval typically surfaces only when a user receives a wrong answer, and the investigation that follows has to reconstruct who indexed it and why, after the fact.

What does source validation actually involve, and what happens if it fails on a given night?

Validation checks whether approved sources remain available and reachable. It does not require rebuilding knowledge stores or reprocessing institutional content, so ongoing operational overhead remains low even as deployments expand. Large university websites often exceed 10,000 pages in total. McGill's current authorized greenlist, 2,010 sources, is a curated subset covering Student Affairs and Services, not the institution's full web footprint. Even at five times that scope, roughly 10,000 sources, comparable to McGill's entire website, a full audit cycle still completes in minutes, because each check is a lightweight URL status request, not a document re-read. That is a fraction of the time a human reviewer would need to manually verify the same list by hand. If a validation run is missed or delayed, the greenlist simply continues operating on its last-confirmed state until the next cycle completes. Detection of a newly broken link is delayed, not the system's ability to operate safely within its existing authorized boundary.

Who has authority to modify the greenlist, versus who can only view it?

During pilot deployments, a single authorized institutional contact typically manages the greenlist directly, in coordination with COMPAiSS, while initial scope is established. This is a pilot-stage practice, not an architectural constraint. The underlying platform supports role separation for institutions that require it: staff who can add, remove, or approve sources, and reviewers, procurement officers, legal counsel, auditors, who need visibility into the current list and its complete change history without modification rights. As a deployment matures, greenlist authority transitions to whatever internal approval structure the institution already uses to publish official content.

COMPAiSS does not create a new governance hierarchy. It attaches an audit trail to the one that already exists.

Isn't a curated list just a smaller version of the same trust problem?

A curated list is a smaller authorized evidence boundary, and that reduction is the governance improvement being claimed, not a restatement of the same problem at smaller scale. In conventional retrieval, the boundary is every document a crawler or connector can technically reach, whether or not anyone reviewed it for AI use. In a greenlist deployment, the boundary is exactly the set of sources a named institutional authority explicitly approved. That does not eliminate the underlying requirement that approved content be accurate and current, no architecture eliminates that. What it changes is whether a failure of that requirement is visible and attributable, or invisible until it reaches a user.


Aligned with How Institutional Web Is Evolving

The shift from navigation to direct AI-mediated answers is already underway.

Many universities and public institutions are currently undergoing significant web restructuring - moving away from large, decentralized site ecosystems toward smaller sets of high-value, authoritative content. At the same time, user behavior is shifting away from traditional site navigation toward direct, AI-mediated question answering.

This creates a specific challenge. During web migration periods, URLs change, content becomes fragmented, and some pages degrade or disappear. Institutions that have invested in retrieval-based AI systems tied to their existing web structure face a compounding problem: the content layer their AI depends on becomes unstable precisely when accurate answers matter most.

COMPAiSS is designed for this environment. Rather than relying on site structure or navigation, it identifies and prioritizes authoritative institutional content algorithmically - the same high-value policies, registrar information, student services guidance, and academic rules that institutions are themselves trying to surface. It filters out low-value or non-governed pages and delivers direct, verified answers without requiring users to navigate complex or transitioning web systems.

As institutions restructure their web presence and users increasingly expect direct AI-mediated answers, COMPAiSS provides a stability layer over an evolving information environment - one that maintains accuracy and institutional control regardless of how the underlying web infrastructure changes.

Institutions can maintain the quality of AI-mediated answers even as their web infrastructure transitions, without rebuilding the AI layer every time content moves.

COMPAiSS abstracts away web complexity. It delivers governed answers from authoritative content - regardless of how that content is structured, hosted, or reorganized.

COMPAiSS and Your Existing Systems

COMPAiSS does not replace your existing workflow systems. It adds a governed intelligence layer they were never designed to provide.

RAG and COMPAiSS address different problems under different circumstances. RAG improves generation quality by connecting the model to relevant documents before a response is produced. It is a meaningful improvement over ungrounded AI, but residual hallucination rates remain a documented challenge even under optimal conditions. For institutions where those residual risks are acceptable, RAG is a reasonable architecture. For institutions where they are not, a different approach is required.

COMPAiSS addresses the authorization problem rather than the retrieval problem. When an institution already knows which sources are authoritative and the scope of permitted responses must be bounded, the relevant question is not which content is most relevant to retrieve. It is whether the system is permitted to generate a response at all. COMPAiSS enforces that authorization before inference occurs. For institutions evaluating AI for the first time, this provides a complete governance architecture without the infrastructure overhead RAG requires. For institutions that have already invested in RAG for other purposes, COMPAiSS addresses a different and complementary requirement.

In both cases, COMPAiSS operates alongside existing CRM and workflow systems rather than replacing them. Existing systems handle process, communication, and record management. COMPAiSS handles governed intelligence: accurate, policy-aligned, institutionally authorized answers.

Your existing CRM / workflow system
System of Record
  • Manages applications and student data
  • Handles communications at scale
  • Tracks workflows and processes
  • Supports recruitment and events
  • Moves students through admission steps
+
COMPAiSS
Governed Intelligence Layer
  • Answers complex policy questions accurately
  • Eliminates advisor-to-advisor variation
  • Handles edge cases and conditional rules
  • Activates internal institutional knowledge
  • Generates ready-to-use, policy-aligned responses
Your CRM manages interactions. COMPAiSS governs the intelligence behind those interactions.

This model also opens access to institutional knowledge that currently exists only in internal documents - admissions guidance, internal policy interpretations, staff-facing materials - that is not publicly available but directly informs how staff use their systems. COMPAiSS can incorporate this content under full institutional control, creating a staff-mode layer that activates institutional knowledge without exposing it publicly.

This is about unlocking institutional knowledge - not CRM data. Nothing enters the system unless the institution explicitly authorizes it.

Low-Risk Evaluation and Approval

A system designed to be tried before it is committed to.

COMPAiSS can be deployed independently, evaluated safely, and discontinued at any time without impacting existing systems. No existing workflows are disrupted. No data is touched. No architecture is modified.

Because it operates without system integration and typically falls below standard procurement thresholds for enterprise software, it can be assessed quickly - without extended implementation cycles, IT project approvals, or operational risk.

An evaluation can begin with a curated set of high-impact policy and admissions content, then expand over time as confidence grows - adding internal documents, additional policy areas, or workflow integration at the institution's pace and under the institution's control.

Start Small

Begin with a defined set of policy content and high-volume questions. No infrastructure commitment required to evaluate real performance against real institutional questions.

Expand at Your Pace

Add internal documents, additional policy areas, or workflow integration over time. Each expansion is institution-controlled and does not require system changes.

Exit Without Consequence

COMPAiSS operates independently of existing systems. Discontinuing it leaves nothing behind - no data dependencies, no integration to unwind, no infrastructure to decommission.

The core procurement argument: COMPAiSS does not ask institutions to commit to an architectural change before they have seen it work. It asks them to evaluate it against their own questions, using their own content, without disrupting anything they currently rely on.

The risk of evaluating it is low. The risk of not evaluating it - while the accuracy and governance problems it addresses continue to compound - is not.

We don't need to rebuild or integrate into existing systems to deliver value. COMPAiSS sits alongside them, reducing cost, complexity, and risk while immediately improving information quality.

Why COMPAiSS Can Be Deployed with Lower Operational Costs

COMPAiSS does not reduce costs by reducing governance, security, or institutional oversight.

Rather, its authorization-first architecture eliminates several infrastructure components commonly required by inference-first RAG deployments, including vector databases, embedding pipelines, synchronization services, retrieval optimization, and ongoing index management.

Lower operational costs are therefore a consequence of architectural simplification, not reduced institutional capability.

Higher Education Institutional Pricing

Transparent, all-inclusive licensing for medium-size universities.

COMPAiSS is offered as an institutional platform license. All infrastructure, AI processing, governance oversight, monitoring, and reporting are included within the annual license fee. Institutions do not manage or pay for underlying components separately.

The pricing below is modelled on a medium-size university deployment (approximately 5,000-8,000 students) with an initial scope of Student Affairs & Services. Year 1 is conservatively projected at approximately 12,000 annual sessions.

Annual License (Y1)
$30,000
Includes $4,000 pilot & implementation fee
Annual License (Y2)
$26,000
All-inclusive
Annual License (Y3)
$27,000
~20% usage growth
Deployment Timeline
2-3 wks
From contract execution

What the license includes: AI processing capacity, secure hosting infrastructure (24/7 uptime, encryption, traffic protection), governance oversight and policy validation (ongoing greenlist maintenance, scope boundary updates, quarterly risk reviews), monitoring and analytics (engagement analytics, administrative dashboards, performance diagnostics), and a staff pilot period of up to 30 days before full student deployment.

The Greenlist Management Dashboard

Every COMPAiSS deployment includes a purpose-built administrative dashboard giving designated institutional staff direct, authenticated control over the approved source boundary. Full governance details, including how large greenlists scale by department and how the greenlist functions as audit evidence, are available here. Administrators can add or remove authorized content sources, including HTML pages, PDFs, policies, procedures, handbooks, research documents, and other institutionally approved materials. Staff can review the complete version history of every modification and commit changes that take effect within minutes. No developer involvement is required for routine governance updates. The example below shows the McGill University greenlist at 2,010 active authorized sources, all verified and current.

✦ COMPAiSS Greenlist Management Dashboard — McGill University
Total Sources
2,010
Active
2,010
Institution
McGill
Unsaved Changes
None
+ Add URL Commit Changes ↺ Refresh Search sources...
StatusSource URL
● Active https://www.mcgill.ca/student-accounts/tuition-fees/general-tuition-and-fees-information/tuition-fees-2026-27
● Active https://www.mcgill.ca/student-accounts/files/student-accounts/new_cnrq_ugrad_admitted_for_2026-27_1.pdf
● Active https://www.mcgill.ca/student-accounts/files/student-accounts/new_cnrq_grad_g1-non-thesis_qualifying_admitted_for_2026-27_3.pdf
● Active https://www.mcgill.ca/student-accounts/files/student-accounts/new_cnrq_grad_post-dent_admitted_for_2026-27.pdf
+ 2,005 more All active, verified and current as of deployment

Changes take effect within minutes of being committed. Every modification is logged in full version history. No developer involvement is required for routine scope updates.

Usage & Growth Framework

Usage LevelInteraction RangeCost ImpactAction
Within Projected Band~12,000/yrFully includedNo action required
Moderate Overuse120-150% of projectionIncremental AI cost onlyMonitored; absorbed where reasonable
Significant Overuse>150% of projectionScaling reviewPricing adjustment discussion
Significant Underuse<50% of projectionLower variable costYear 1 review may be conducted

Procurement & Security Readiness

CapabilityStatus
Commercial deploymentActive
Institutional pilotsActive
HECVAT packageComplete
AI governance documentationComplete
Privacy documentationComplete
Canadian hostingOperational
Procurement packageAvailable

COMPAiSS has been prepared for institutional procurement from the outset. Supporting documentation is available for institutional procurement and security review.

Available Documentation

HECVAT 4.1.6
AI Governance Policy
System Architecture & Data Flow
Privacy Policy
Data Privacy Impact Assessment
Personal Data Inventory
Access Control Policy
Incident Response Plan
Business Continuity Plan
Disaster Recovery Plan
Employee Onboarding & Offboarding Policy
Accessibility Conformance Statement
Internal Audit Process

Enterprise Security Controls

Canadian-only infrastructure
Azure OpenAI Canada East inference
Fly.io Toronto hosting
Cloudflare security services
TLS 1.2+ encryption in transit
AES-256 encryption at rest
Role-Based Access Control (RBAC)
Multi-Factor Authentication
Least-privilege access management
Institution-controlled authorization boundary (greenlist)
Institution-controlled deployment management
Documented incident response procedures
Business continuity planning, 4-hour RTO target
Disaster recovery procedures, 24-hour RPO target
Annual governance reviews
Operational audit logging

Framework Alignment

Canada's Voluntary Code of Conduct on Responsible Generative AI
HECVAT 4.1.6
NIST AI Risk Management Framework
NIST Cybersecurity Framework
PIPEDA
Provincial privacy legislation
WCAG 2.1 AA (roadmap and conformance documentation)

Privacy by Design

Queries are processed statelessly
Conversation history is not retained
Institutional data is not used for model training
Institutional deployments remain isolated from one another
No cross-institutional knowledge sharing
Institutions retain complete authority over authorized knowledge sources

Deployment Support

Every institutional deployment includes:

Greenlist development assistance
Initial validation and testing
Administrator onboarding
Deployment support
Governance consultation
Post-deployment refinement

Transparent Compliance

Institutional trust is built through transparent disclosure, not perfect checklists. Our procurement documentation identifies both implemented controls and those currently under development, allowing institutions to evaluate the platform against an accurate, current compliance posture rather than aspirational claims.

Institutional Procurement Package

Includes HECVAT 4.1.6, AI Governance documentation, Incident Response Plan, Business Continuity Plan, Disaster Recovery Plan, DPIA, Architecture documentation, and supporting procurement materials.

Get the Procurement Package →

Competitive Comparison - Higher Education

Annual licensing costs for comparable AI assistant platforms deployed in higher education. Figures exclude implementation, onboarding, and governance configuration - which typically add 15-30% to first-year total cost of ownership for competing solutions.

PlatformSmall UniversityMedium UniversityLarge UniversityPricing Model
Bolt AI Assistants$30K-$50K$50K-$80K$80K-$110KAnnual subscription
Capacity AI$30K-$55K$55K-$85K$85K-$120KUsage + platform fee
Coveo AI$40K-$70K$70K-$100K$100K-$140KEnterprise contract
Intercom (Education)$45K-$75K$75K-$110K$110K-$150KUsage + seats
Ivy.ai$45K-$75K$75K-$110K$110K-$145KAnnual subscription
Mainstay (AdmitHub)$35K-$60K$60K-$95K$95K-$135KAnnual subscription
Ocelot$20K-$40K$40K-$65K$65K-$90KAnnual subscription
Verge AI$25K-$45K$45K-$75K$75K-$120KEnterprise contract
Zendesk Answer Bot$25K-$45K$45K-$75K$75K-$105KPer-agent pricing
COMPAiSS (Medium Univ.) - Y1: $30,000
Y2: $26,000
Y3: $27,000
- Governance-first institutional pilot - implementation fee, 6-month review, all-inclusive

1. Competitor pricing reflects publicly available market estimates and analyst benchmarks. All vendors use custom quote-based models; figures represent typical institutional ranges for a medium-size university deployment.

2. Pricing reflects a medium-size institutional deployment scoped to approximately 5,000–8,000 full-time equivalent students with an initial Student Affairs & Services mandate projected at approximately 12,000 annual sessions. Institutions outside this range are quoted separately based on deployment scope, student population, and projected session volume.

3. COMPAiSS is sized for the specific deployment scope. The cost advantage would narrow at enterprise scale; however, COMPAiSS's pre-inference governance architecture remains unique across the competitive set regardless of scale.

4. Platforms in this comparison fall into two architectural categories: FAQ-based systems that match queries to pre-authored answer libraries, and inference-first chatbots that generate responses from underlying AI models without a prior authorization layer. COMPAiSS operates in neither category. Its execution gate means inference does not occur until pre-generation authorization is satisfied — a structural distinction from all identified competitors, which uniformly treat governance as a post-generation filter or content restriction layer.

🏗

Pre-inference governance

Source validation and scope enforcement occur before AI generation. No competitor operates at this architectural layer. Every platform above applies filtering after inference - meaning hallucinated content is produced first and suppressed second.

🔌

Minimal IT integration

Embeds directly within the institution's existing website. Optional SSO requires a single app registration in Azure AD - no server installations, database migrations, or infrastructure changes required.

📋

Procurement-ready

Pre-inference authorization means every interaction follows a documented, reproducible governance path. Compatible with AI governance panel requirements, AIA frameworks, and government procurement standards.


Return on Investment

Adjust the sliders to match your institution and see your estimated annual return.

Queries diverted
8,750
per year
Staff time recovered
$70,000
estimated annual value
Net annual return
$40,000
2.3x return on investment

Industry reference benchmarks — Canadian universities

STAFF COST PER INQUIRY
$5 – $9 CAD
Based on 8 to 12 routine inquiries per hour, at a fully loaded salary of $55,000 to $70,000 CAD per year.
ANNUAL INQUIRY VOLUME
8,000 – 40,000
8,000 to 15,000 for smaller institutions; 20,000 to 40,000 for medium-size universities. The University of Toronto Registrar reported handling over 120,000 student inquiries in 2024 across admissions, financial aid, fees, and transcripts.
QUERY DIVERSION RATE
25% – 55%
25 to 35% in early deployments; 40 to 55% at deployment maturity.

Healthcare & Hospital Systems

Bridging the gap between limited FAQ bots and ungoverned general-purpose AI.

Hospitals today face two competing problems. On one side, patients, families, and visitors increasingly turn to general-purpose AI systems, ChatGPT, Gemini, Copilot, Claude, Perplexity, for answers about hospital services, surgery preparation, patient education, referrals, and visiting policies. Those systems were not designed to answer exclusively from a hospital's authorized information, and can produce responses derived from general training knowledge or sources the institution has never approved.

On the other side, traditional FAQ bots are safe precisely because they only answer a limited, pre-scripted list of questions, but that same limitation leaves many users unable to find what they need, pushing them back toward general-purpose AI anyway.

COMPAiSS was built for the space between them: broader and more capable than a scripted FAQ tool, but permanently bounded to only the hospital's own authorized sources, never general training knowledge.

Rather than generating an answer and attempting to constrain it afterward, COMPAiSS authorizes institutional content before inference begins. If an answer cannot be supported from authorized hospital sources, COMPAiSS does not answer from outside those sources.

Case Study: MUHC and the New Governance Gap in Healthcare AI

COMPAiSS's healthcare architecture was tested against a real, publicly documented case: the McGill University Health Centre. Measured traffic to muhc.ca declined by an estimated 16,000 monthly visits in a single month, a pattern independently corroborated at a second, unrelated Canadian hospital, IWK Health Centre in Halifax, whose organic search traffic fell roughly 35% over six consecutive months.

Live testing against real patient questions found that four widely used general-purpose AI systems, and Google's own AI Overview, will confidently answer individualized clinical questions with no connection to any hospital's authorized guidance. The case study documents this evidence directly, alongside the governance architecture built to close the gap.

Read the Full Healthcare Governance Case Study →

Typical Healthcare Applications

Healthcare organizations deploy COMPAiSS to support a broad range of public information services, including:

  • Patient education and health information
  • Surgery preparation and recovery guidance
  • Hospital maps, parking, and wayfinding
  • Clinic, department, and physician information
  • Visiting hours and visitor policies
  • Careers and recruitment
  • Research and academic programs
  • Institutional policies and public notices
  • Foundation and community information

No electronic health record integration is required. No patient records are accessed. Existing hospital websites remain the authoritative source of institutional information.

Institutional Benefits

🛡

Safer Public-Facing AI

Provides answers exclusively from hospital-authorized information rather than general AI training knowledge.

Reduces Reliance on General AI

Offers patients and families a trusted institutional alternative to general-purpose AI for hospital-specific questions.

🔍

Improves Access to Existing Information

Helps users find patient education, surgery guides, and hospital services through natural conversation rather than website navigation alone.

Reduces Routine Enquiries

Provides immediate answers to many repetitive public enquiries, helping reduce demand on switchboards and information lines.

👥

Supports Call-Centre Agents Directly

Agents face the same pressure patients do: an answer needed quickly, with an authoritative-sounding source close at hand. Without a governed option, staff searching for something they don't already know are not immune to the same ungoverned general AI tools driving the patient-side problem. COMPAiSS gives agents a reliable, governed first stop instead, which matters more than it might for patient self-service: an agent's answer is delivered as the hospital's institutional voice, not a private search a patient did on their own.

Preserves Institutional Authority

Every institutional response remains grounded in hospital-approved content under the organization's control.

📋

Complete Auditability

Every answer can be traced back to authorized institutional sources, simplifying review, correction, and governance.

🌐

Bilingual and Multilingual Support

Supports public communication in the languages your patient population actually needs, governed the same way regardless of language.

💬

Consistent Information Across Channels

The same governed knowledge base serves patients, families, visitors, referring providers, and staff, so the answer doesn't change depending on who's asking or which channel they used.

Typical Deployment Profiles

Healthcare deployments are individually scoped according to organizational complexity and public information requirements.

OrganizationTypical Deployment
Community HospitalSingle-site public information, patient education, visitor services
Regional Hospital NetworkMulti-site deployment with shared patient education, multilingual services, and centralized governance
Academic Health CentreLarge patient education library, research, multiple hospitals, multilingual public information
Provincial / Regional Health AuthorityEnterprise deployment across multiple hospital systems and health services

Governance Impact

Healthcare organizations are increasingly competing with general-purpose AI as the public's first source of information. A Governance Impact review looks at public website traffic, institutional information complexity, and the likelihood that users are seeking answers from general AI rather than authorized hospital sources.

This assessment is intended to quantify governance exposure, not financial return. It answers "how exposed is this institution," not "how much does this save." See Operational Impact below for the dollar-based estimate.

Governance Exposure

Request a governance impact review scoped to your institution's public traffic and content footprint.

Request a Governance Impact Review →

Operational Impact & ROI

Adjust the sliders to match your hospital's call volume and see your estimated annual return. This calculator is driven by call-centre activity, not website traffic, see the note below.

Calls deflected
125,000
per year
Savings per interaction
$7.98
staffed cost minus AI cost
Estimated annual savings
$997,500
400× cost reduction per interaction

Industry reference benchmarks — healthcare self-service

ASSISTED VS. SELF-SERVICE COST
$8–12 vs. $0.10–0.25
Gartner: live/assisted contact channels average $8–12 CAD per interaction; self-service interactions run $0.10–0.25.
REAL GOVERNMENT DEPLOYMENT
4.81× ROI
Bernalillo County, NM: AI-handled interactions at $0.99 vs. $4.59 for human-handled, measured over an actual 18-month deployment.
HOSPITAL SWITCHBOARD VOLUME
200K – 800K
Typical annual call range for a single hospital switchboard. Multi-site networks and academic health centres run higher, see deployment profiles above.

This calculator estimates operational impact based on call-centre activity, not website traffic. Website traffic is a governance-exposure signal (see Governance Impact above), not a direct input to this dollar estimate, the two measure different things and are kept separate deliberately. All figures are estimates for planning purposes; final figures are established through a scoped assessment.




Return on Investment - Call Centre & Public Sector

Modelled on the live Service Canada beta. Adjust the sliders to match your call centre and see your estimated annual savings.

● Live beta — compaiss.ca/service-canada
Load a profile:
Total annual calls
8.0M
current call volume
Savings per interaction
$8.31
staffed cost minus AI cost
Annual savings at 30% deflection
$19.94M
416× cost reduction
Cost per interaction — staffed vs COMPAiSS
Staffed call COMPAiSS
Staffed
$8.33
$8.33 / call
COMPAiSS
$0.02 / call
Annual savings by deflection rate — Service Canada

Each row shows annual savings if that percentage of total calls is handled by COMPAiSS instead of staff.

At Service Canada's confirmed interaction cost of $0.02 — versus $8.33 per staffed call — COMPAiSS delivers a 416× cost reduction per interaction. Deflecting 30% of total calls saves an estimated $19.94M annually, while staff remain fully available for complex cases, appeals, and sensitive client interactions.

Industry reference benchmarks — Canadian public sector call centres

Cost per staffed call
$6 – $12 CAD
Federal and provincial call centres typically run $7–$10 per interaction fully loaded; Service Canada confirmed at $8.33 (2023–24).
Annual call volume
300K – 20M
Service Canada handles ~8M calls annually across EI, CPP, OAS and GIS. Municipal 311 centres average 300K–2M. Hospital switchboards 200K–800K.
Deflection potential
25% – 65%
Informational queries — eligibility, status checks, hours, locations, program details — typically represent 50–65% of inbound volume at public sector centres and are candidates for AI deflection.

Service Canada defaults: ~8M annual calls across EI, CPP, OAS and other programs at ~$8.33 per staffed interaction (2023–24 figures). COMPAiSS interaction cost confirmed at $0.02. Savings projections assume COMPAiSS handles calls that would otherwise be answered by staff — complex cases, appeals, and sensitive interactions remain with trained staff. All figures are estimates for planning purposes.

Government AI Evaluation: CRA GenAI Chatbot vs COMPAiSS

In June 2026, COMPAiSS was evaluated against the Canada Revenue Agency GenAI Chatbot using twelve tax and benefits scenarios covering filing deadlines, taxpayer rights, government programs, multilingual access, temporal accuracy, source governance, and public-service accountability.

The evaluation identified expired guidance, omitted taxpayer protections, missing federal program references, limited multilingual service capability, and differences in source transparency and temporal accuracy.

The findings also relate directly to Canada's National AI Strategy, including its emphasis on protecting Canadians, safeguarding trust, empowering Canadians to benefit from AI, and strengthening public institutions through responsible AI adoption.

Read the Full CRA Evaluation Report →

See the Canada AI Strategy discussion in the AI Risk Assessment Framework →

Government Procurement Activity

COMPAiSS is publicly listed on the Government of Canada's CanadaBuys platform as an interested supplier for the Artificial Intelligence Source List procurement opportunity.

The CanadaBuys profile includes a direct link to COMPAiSS and a description of the execution-gated governance architecture developed for regulated institutional environments.

View CanadaBuys Supplier Profile →

Government of Canada AI Answers — Published Trial Results

The Canada.ca Experience Office published results in December 2025 from enterprise-scale AI trials across ESDC, CRA, IRCC, and seven other federal institutions. Across 7,851 questions and three trials, the system achieved accuracy rates of 94 to 96.7 percent sourcing answers exclusively from Government of Canada websites. Every response carried the disclaimer: "AI can make mistakes, always check your answer."

A 94 to 96 percent accuracy rate is a genuine achievement. It also means that across 7,851 trial questions, between 235 and 471 responses contained errors - delivered to Canadians seeking help with EI, CPP, tax accounts, and immigration status. Scaled to Service Canada's confirmed volume of approximately 8 million annual interactions, a 3 to 6 percent residual error rate represents between 240,000 and 480,000 incorrect responses per year. The disclaimer is the governance layer for those errors. COMPAiSS addresses the same challenge differently: the pre-inference authorization gate prevents generation from occurring when verified evidence is unavailable, rather than disclosing after the fact that generation may have been incorrect.

Read the Government of Canada AI Answers trial results

Schedule an Institutional Demonstration

COMPAiSS is available for institutional evaluation. To schedule a demonstration, discuss a pilot deployment, or request governance documentation for procurement review:

Request a Demonstration →