Artificial Intelligence in the Sectional Title Industry: A Powerful Tool, Not an Oracle (By Taylor Noble)

The promise, and the peril, of AI-generated information in Community Scheme management.

Artificial intelligence has arrived in the sectional title industry, and it is not leaving. Managing Agents are using it to draft correspondence, summarise financial statements and prepare meeting packs in a fraction of the time. Trustees are using it to make sense of dense legislation. Owners are using it to understand their rights before an AGM. Used well, AI is one of the most significant productivity tools our industry has seen in decades.

But alongside the genuine benefits, a quieter problem is growing — and it deserves the industry’s honest attention. Increasingly, Managing Agents, Trustees, Portfolio Managers and owners are treating AI output as established fact. Legal opinions are being formed, disputes escalated and trustee resolutions passed on the strength of information that was never verified against the Sectional Titles Schemes Management Act, the scheme’s registered rules, or any authoritative source at all.

This article looks at both sides: what AI does well in our space, where it goes wrong, and what responsible use should look like for everyone involved in Community Schemes.

Where AI Genuinely Adds Value
It is important to start here, because the answer to AI misuse is not to abandon the technology — it is to use it properly. In sectional title administration, AI is already delivering real value in areas such as:

• Administrative efficiency. Drafting owner correspondence, converting rough meeting notes into structured minutes, preparing first drafts of AGM and SGM packs, and summarising lengthy documents. Work that once took hours can be completed in minutes, freeing Portfolio Managers to focus on judgement-based work.
• Financial analysis and reporting. AI tools can interrogate levy rolls, flag arrears trends, summarise management accounts and translate financial statements into plain language that owners actually understand — a long-standing communication gap in our industry.
• Accessibility of legislation. The STSMA, the Prescribed Management and Conduct Rules, and CSOS practice directives are not light reading. AI can explain these frameworks in accessible language, helping Trustees and owners engage more meaningfully with scheme governance.
• Communication at scale. Triaging high-volume scheme inboxes, drafting responses to routine queries, and producing consistent, professional communication across a large portfolio.

None of this should be understated. Schemes benefit when their Managing Agents are more efficient, their Trustees better informed, and their owners better able to understand governance. The problem is not the tool. The problem is how the tool is being used.

The Core Risk: Fluent, Confident — and Sometimes Wrong
Large language models are designed to produce fluent, confident, plausible-sounding text. They are not designed to guarantee accuracy, and they do not know when they are wrong. In AI terminology this is called “hallucination”: the system generates information that sounds authoritative but is partly or entirely fabricated.
In most industries a hallucination is an inconvenience. In sectional title, it can be a legal and financial problem, because our industry runs on precise statutory mechanics: notice periods, quorum requirements, resolution thresholds, fiduciary duties and prescribed procedures. Get the detail wrong, and decisions can be invalid, disputes can escalate, and trustees can be exposed.

Consider the failure patterns already emerging in practice:

• Invented or misquoted legal provisions. AI tools will confidently cite a “Prescribed Management Rule” that does not exist, attribute a requirement to the wrong section of the STSMA, or blend the old Sectional Titles Act regime with the current STSMA framework. An owner armed with a fabricated rule number can derail an entire general meeting — and a trustee who accepts it without checking can compound the error.
• Jurisdictional confusion. Much of the material AI models are trained on is American, British or Australian. Concepts from strata title (Australia), condominium law (USA) or leasehold (UK) quietly bleed into answers about South African sectional title. Advice about “HOA boards”, “special assessments” or foreign dispute processes may sound relevant, but it does not reflect the STSMA, the CSOS Act or our regulatory environment. A homeowners association governed by a constitution and a body corporate governed by the STSMA are fundamentally different legal creatures — a distinction AI frequently blurs.
• Outdated information presented as current. CSOS practice directives, levy contribution mechanics, fidelity insurance requirements and case law develop over time. An AI answer may reflect the position as it stood years ago, presented with complete confidence as the current state of the law.
• Scheme-specific blindness. AI does not know your scheme’s registered conduct and management rules, its unique exclusive use arrangements, or the resolutions in its minute book. Generic answers about “what the rules say” are meaningless when a scheme has properly registered amended rules — yet owners routinely present generic AI output as though it overrides the scheme’s own registered documentation.
• Garbage in, garbage out. The quality of AI output is heavily dependent on the quality of the prompt. A vague, one-line question produces a vague, generic answer. An owner who asks “can trustees do this?” without context, documents or jurisdiction will receive an answer that may be confidently wrong on every material point — and then circulate it in the scheme’s WhatsApp group as fact.

The Misinformation Multiplier
That last point deserves emphasis, because it is where individual error becomes industry-wide risk. Community Schemes are, by nature, communities — with group chats, notice boards and AGMs. A single piece of confident AI-generated misinformation (“the Act says trustees cannot raise a special levy without an owners’ vote”, for example) can spread through a scheme in hours, harden into accepted “fact”, and poison the relationship between owners, Trustees and the Managing Agent.

Managing Agents are increasingly finding themselves not just administering schemes, but correcting the record — responding to demands, objections and even CSOS applications built on AI-generated interpretations that were never checked against the actual legislation or the scheme’s registered rules. This consumes time, inflames conflict and erodes trust on all sides.

There is also a professional dimension. A Portfolio Manager or Trustee who acts on unverified AI output is not absolved of responsibility because “the AI said so”. Trustees owe fiduciary duties to the body corporate. Managing Agents owe professional duties to their clients. AI does not carry professional indemnity insurance — the human who relies on it does.

What Responsible AI Use Looks Like
The industry should neither ban AI nor blindly embrace it. Instead, we suggest a simple framework for all stakeholders:

1. Treat AI output as a first draft, never a final answer. Use it to structure thinking, summarise documents and produce drafts — then verify every legal or financial assertion against the primary source: the STSMA itself, the Prescribed Rules, the scheme’s registered rules, CSOS directives, or a qualified professional.
2. Prompt with context, documents and jurisdiction. The difference between a dangerous answer and a useful one is often the prompt. Specify South African law. Specify sectional title (not HOA, not strata). Where appropriate, provide the actual rule or document you are asking about, rather than asking the AI to recall it from memory — recall is where fabrication happens.
3. Demand sources — and check them. If an AI cites a rule, a section or a case, look it up. If it cannot be found in the actual legislation or a reputable source, it does not exist, no matter how confident the answer sounded.
4. Never substitute AI for professional advice on material decisions. Special levies, rule amendments, extensions of schemes, legal proceedings, insurance and employment matters warrant advice from attorneys, auditors or experienced practitioners. AI can help you prepare better questions for those professionals — it cannot replace them.
5. Managing agents: adopt internal AI policies. Firms using AI in their workflows should define where it may be used, what must be human-reviewed before it leaves the building, and how client and owner data is protected when using AI tools. Staff should be trained not only in how to use AI, but in how it fails.
6. Trustees and owners: verify before you circulate. Before an AI-generated interpretation is posted to the scheme’s group or raised at a meeting as fact, it should be checked — ideally with the managing agent or the scheme’s professional advisors. Being first to share is not the same as being right.

The Way Forward
AI is going to make good managing agents better, well-run schemes more efficient, and legislation more accessible to ordinary owners. Those are outcomes worth pursuing, and our industry should pursue them openly rather than pretending the technology away.

But sectional title is a field where precision matters — where a notice period, a resolution threshold or a single rule number can determine whether a decision stands or falls. In that environment, unverified information is not neutral; it is a liability. The industry’s message to every stakeholder should be consistent and simple:

Use AI to work faster. Never use it to think less. And never mistake confidence for correctness.

Use AI to work faster. Never use it to think less.

And never mistake confidence for correctness.

The Act, the registered rules and qualified professionals remain the authorities in this industry. AI, used responsibly, is a remarkable assistant to all three — and a poor substitute for any of them.