When you contact your condominium manager, you deserve a thoughtful answer from someone who understands your property and the issue you have raised. When you use a system that handles your information, you deserve care in how it is designed, tested, and maintained.
At MVP Condominium Property Management, those expectations guide our use of artificial intelligence.
AI can help us write more clearly, find relevant information, and develop useful tools. Used well, it can give our people more time to focus on the communities they serve. Those benefits depend on people applying their knowledge, checking the work, and accepting responsibility for the result.
Our standard is human judgment at the beginning, human oversight throughout, and human approval before the work reaches you.
This policy establishes separate requirements for business communication and software development. It applies to employees and anyone using AI to communicate or develop systems on MVP's behalf.
Communication: people decide what the message says
For communication, the employee must determine the substance of the response before asking AI to refine it.
That means understanding the question, checking the relevant facts, deciding what needs to be said, and considering what action is reasonable. The employee must prepare their own draft or substantive notes that express those decisions.
AI may then help with spelling, grammar, punctuation, clarity, organization, tone, and length. It can remove repetition, soften an unnecessarily harsh sentence, and make a complicated explanation easier to follow. The final message must preserve the employee's intended meaning and natural voice.
Pasting an incoming email into AI, asking it to decide how to respond, and sending the generated answer does not meet this policy. A quick review before sending does not replace the thinking that should have happened first.
A useful response answers the question and respects the recipient's time. A straightforward issue may need only a few sentences. More complicated matters may require a detailed explanation. Length should follow the needs of the situation.
Requests also need a practical basis. An AI suggestion to ask a contractor to reconcile current findings against a ten-year-old report does not establish that the exercise is necessary or within the contractor's assignment. The employee must understand why the comparison matters and whether that person is responsible for doing it.
Approved AI tools may help locate and summarize records, including declarations, contracts, meeting minutes, and reports. Before relying on the result, the employee must check the original document, its relevant version or amendments, and the surrounding context. An interpretation requiring professional advice must be referred to the appropriate qualified adviser.
Every outgoing message requires the responsible employee to check its facts, requests, conclusions, and commitments. AI must not independently send replies or make commitments on MVP's behalf.
Software development: human direction, testing, and approval
Qualified people must direct, understand, test, and approve the work.
AI can help create an application, website, integration, or automated workflow. A working demonstration shows that a feature can run; it does not establish that the system protects information, handles mistakes correctly, or is ready for everyday use.
Our development requirements place AI assistance within a process led by people:
- People define the work. A responsible person establishes the purpose, requirements, intended users, information involved, and limits on what the system may do. Qualified technical professionals assess the design and potential threats before implementation.
- AI assists with implementation. Developers may use AI to produce code, explore approaches, prepare tests, and support troubleshooting. They must understand the resulting implementation and be capable of reviewing and correcting it.
- People verify and approve the result. Qualified developers and security professionals must review the implementation and oversee functional tests, diagnostics, vulnerability checks, and threat assessments appropriate to the system. Privacy and other applicable requirements must be assessed before release. Asking AI to approve its own output is insufficient.
- People remain responsible after release. Each system must have an identified owner for maintenance, updates, security findings, and incident response. Changes to code, permissions, connected services, or data use must receive appropriate review and testing.
Verification must cover how the system behaves when something goes wrong, as well as when everything works. This includes checking access restrictions, protecting credentials, reviewing outside software components, and confirming recovery arrangements. Findings and release decisions must be documented, with issues that prevent safe use addressed before deployment.
Security review continues throughout a system's working life. NIST's Secure Software Development Framework includes both secure development practices and an ongoing response to vulnerabilities discovered after release. NIST guidance
Qualified oversight may come from internal professionals or appropriately engaged external specialists. An AI development platform's claims do not replace that review.
Privacy is a requirement in both communication and development
Employees must use company-approved AI tools and accounts for business work. Approval must consider the intended use and the information involved. Personal and confidential information must be limited to what is necessary for an authorized purpose.
Before a tool receives that information, the responsible reviewers must assess where it goes, who can access it, how long it is retained, whether it can be used for model training, and how it can be deleted. Appropriate contractual protections and access controls must be established. Canada's privacy regulator provides guidance on assessing these aspects of outside service providers. Privacy Commissioner guidance
AI use must meet applicable privacy laws, including the Personal Information Protection and Electronic Documents Act — PIPEDA — where it applies, together with relevant professional and contractual obligations. Privacy assessments and specialist advice must be obtained where needed. Canadian privacy authorities make clear that existing privacy obligations continue to apply when organizations use or develop generative AI. Canadian privacy principles for generative AI
Guided by Ontario's responsible-AI principles
Ontario's Responsible Use of Artificial Intelligence Directive sets expectations for provincial ministries and agencies. Its principles emphasize useful and proportionate AI use, privacy and safety, fairness and human rights, transparency, and continuing human accountability. Ontario also identifies these principles as a model other organizations can use when developing their own policies. Ontario's trustworthy AI framework
MVP adopts these principles as a guide for its own work: use AI for a clear purpose, assess risks before use, protect information, check for unfair outcomes, explain significant AI-assisted conclusions, and give people a way to reach a responsible person. Human oversight must continue as systems and circumstances change. Ontario's responsible-use directive
This is a voluntary alignment with Ontario's public-sector guidance, not a claim of government approval or certification. It does not replace the privacy, human-rights, professional, contractual, or other legal obligations that apply to MVP.
Accountability in practice
MVP management is responsible for approving uses, assigning accountable owners, and ensuring staff understand these requirements. Suspected errors, inappropriate disclosures, or security weaknesses must be reported promptly so they can be assessed and addressed. This policy must be reviewed as our uses, technology, and applicable requirements change.
People receiving our communications must be able to ask for an explanation, question a conclusion, or raise a concern with a responsible person. If AI-assisted work contributes to an error, we remain responsible for addressing it.
Our aim is to make everyday service clearer, more useful, and more responsive. The judgment behind that service — and the responsibility for it — remains with MVP.