Ask Your HOA's Books a Question, Get a Real Answer: Inside LotWize's AI Financial Q&A

Every HOA board has had this moment. A board member asks, mid-meeting, "how much are we behind on collections this quarter?" or "are we still on budget for landscaping?" The treasurer says "let me check" — and then spends the next ten minutes tabbing between a budget-vs-actual report, an AR aging export, and a spreadsheet of notes to piece together an answer that should have taken ten seconds.

That gap between "the data exists" and "someone can state the number out loud" is where most board meetings lose momentum. LotWize's AI Financial Q&A, live on the board reports page, closes it. A board member types a question in plain English — "how much did we collect this month?" or "which homeowners are 90+ days late?" — and gets back a specific, numbers-first answer pulled from the community's actual ledger, not a generic explanation of how HOA finances typically work.

What the Feature Actually Does

The AI Financial Q&A lives at the top of /board/reports, above the grid of individual report pages (balance sheet, cash flow, trial balance, AR aging, and the rest). It's a single text box with a row of example-question chips underneath it — "How much did we collect this month?", "Are we on budget for landscaping?", "What is our total outstanding receivables?" — that a board member can click to fire off instantly, or ignore in favor of typing their own question.

Under the box, the request goes to a Claude model configured with five specific tools, each one a real database query scoped to the requesting HOA:

• Income summary — assessments collected, late fees, special assessment payments, refunds, and approved expenses by category, for any date range.
• AR aging — outstanding balances per property, bucketed by how many days overdue they are.
• Budget vs. actual — budgeted amount, year-to-date spend, variance, and an "on track / near budget / over budget" status for every expense category.
• Delinquent list — homeowners past the community's configured grace period, sorted by days overdue.
• Assessment roll — every active property's current-period payment status (paid, pending, offline, or no record), with a summary count across the whole roll.

When a board member asks a question, the model decides which of these tools it actually needs, calls them, and writes its answer from what the queries return — it can chain up to five tool calls in a single question if the answer requires combining more than one dataset (for example, "are we collecting enough to cover this month's expenses?" pulls both the income summary and the budget actuals). The answer that comes back cites real figures: "$14,230.00 collected in assessments this month, with $850.00 in late fees" — not an estimate, and not the kind of confident-sounding but ungrounded number a general-purpose chatbot might produce if you pasted your numbers into it.

Why This Is Different From "Just Asking ChatGPT"

It's tempting to assume any AI-answers-my-question feature is just a wrapper around a language model's general knowledge. This one specifically isn't, and that distinction is the whole point.

A general-purpose AI assistant asked "how much are we behind on collections?" has no way to answer that question — it doesn't have your community's data, so at best it can explain what an AR aging report is in the abstract. LotWize's Financial Q&A is built the opposite way: the model has no access to the community's numbers except through the five tools listed above, and its system prompt instructs it to answer using those tools rather than from general knowledge. If a question falls outside what the tools can retrieve — say, a request for legal advice about a lien — the honest answer is that the tool won't manufacture one, because there's no tool call that produces it.

This matters more than it sounds like it should. A treasurer relying on an AI answer to report a number to the board needs that number to be traceable back to the actual ledger, not a plausible-sounding approximation. Because every figure the tool states comes from a live query against assessmentPayment, expense, budget, and property records for that specific organization, the answer is exactly as accurate as the books themselves — which is the correct failure mode. If the underlying data is wrong, the AI will faithfully report the wrong number rather than quietly smoothing it over, and a board that finds a surprising answer knows to go check the source records rather than the AI.

A Real Board Meeting, Before and After

Picture a typical monthly board meeting for a 60-unit, self-managed community. Five questions come up, the kind that come up in nearly every meeting:

1. "How much did we bring in this month?"
2. "Who's furthest behind on dues?"
3. "Are we over budget on any category?"
4. "What's our total outstanding receivables?"
5. "How many units are still unpaid for the current period?"

Without the Q&A tool, the treasurer answers these by opening the income/expense PDF export for question one, the AR aging PDF for questions two and four, the budget-vs-actual report for question three, and the assessment roll for question five — four separate report pages, each requiring a date range selection and a scan for the right row. In a meeting, that's five to ten minutes of "let me pull that up" while the rest of the board waits.

With the Q&A tool, each question is typed or clicked as an example chip, and each answer returns in a few seconds, phrased in a sentence rather than requiring the treasurer to interpret a table live in front of the board. The treasurer isn't slower at reading reports because they lack skill — they're slower because the workflow requires navigating to the right report, setting the right filter, and translating a table into a spoken sentence, five separate times. The Q&A tool collapses that into one interaction per question.

Who Can Ask, and What Guardrails Exist

The Financial Q&A isn't open to every portal user. The endpoint checks the requester's role in the organization and only allows super_admin, board_member, and committee_member roles to use it — a homeowner logged into their own portal account cannot query the community's financial data this way, even indirectly. That access control exists independently of and in addition to the plan gate: the feature sits behind hasFinancialReports, which is available starting on LotWize's Growth plan, consistent with the rest of the financial reporting suite (balance sheet, cash flow, trial balance, AP/AR aging, and 1099s) living behind the same gate.

There's also a straightforward rate limit — 20 questions per hour per user — which exists for the same reason the free Document Analyzer tool caps its usage: every question is a real model call with a real cost, and the limit is generous enough to cover an actual board meeting's worth of back-and-forth without being an invitation to script against the endpoint. In practice, a board burns through two or three questions in a typical meeting, not twenty.

Why This Matters More for Self-Managed Boards Than It Sounds

Professional management companies field this exact kind of question constantly — a board member calls the property manager and asks for a number, and the property manager (or their staff accountant) either already has it memorized or can pull it quickly because reporting is their full-time job. Self-managed boards don't have that layer. The treasurer is a volunteer who might look at the books once a week, not once an hour, and the five-report shuffle described above is the tax that volunteer status carries — every "quick question" costs real, uncompensated time.

The Financial Q&A doesn't replace the underlying reports — a board still needs the full balance sheet for an audit, the trial balance for year-end close, the AR aging export to actually chase down who owes what. What it replaces is the specific pattern of "someone asks a simple factual question, and answering it correctly requires navigating a multi-page reporting UI." That pattern shows up constantly in board meetings, in board member group chats, and in the treasurer's own head when they're trying to sanity-check a budget line before a vote. Turning it into a text box instead of a report-navigation exercise is a small UX change with an outsized effect on how much friction a volunteer board has to push through to make an informed decision.

It's also a natural complement to LotWize's AI bookkeeping automation, which handles the categorization side of the books, and the wider financial reporting suite that the Q&A tool sits on top of. Clean, correctly categorized books make the Q&A tool's answers more useful — garbage in, garbage out applies here exactly as it would to a human reading the same reports.

Key Takeaways

LotWize's AI Financial Q&A, on /board/reports, answers plain-English financial questions by calling five real database tools — income summary, AR aging, budget vs. actual, delinquent list, and assessment roll — not by generating a plausible-sounding guess.

Every number in an answer traces back to a live query against the community's actual ledger, so the AI is exactly as accurate as the underlying books — it reports data faithfully rather than smoothing over bad data.

Access is restricted to board members, committee members, and super admins (not homeowners), sits behind the Growth-plan hasFinancialReports gate, and is rate-limited to 20 questions per hour per user.

A meeting that would otherwise require navigating four or five separate report pages to answer routine questions collapses into a single text box, with each answer returning in a few seconds.

The tool doesn't replace formal reports like the balance sheet or trial balance — it replaces the specific friction of answering simple factual questions on the spot during a meeting or a board discussion.

If your board is still tabbing between report tabs to answer "are we on budget?" mid-meeting, start a free LotWize trial and try asking your own books a question. For the categorization work that keeps those answers accurate in the first place, see how AI bookkeeping automation turns month-end close into a 30-minute review instead of a half-day data-entry marathon.