Brickwise DocsGo to app

Limits & Expectations

Brickwise is built around one rule: be honest about what's real and what isn't. This page collects every constraint, gap, and known rough edge worth knowing about before you rely on the app — stated plainly, not buried in a support article.

What the AI features will and won't do

Brickwise's AI features (Ezrah, Due Diligence, Investment Manager, and listing extraction) share hard rules:

  • No AI feature invents a financial number. Every dollar figure, percentage, or ratio comes from a deterministic calculator (the same math the forms use) or from data already stored — never guessed. See the Glossary for the VERIFIED FACT / CALCULATED RESULT / ESTIMATE / AI INTERPRETATION labeling system Ezrah uses to make this traceable.
  • The Investment Score's recommendation is never AI-chosen. Investment Manager can narrate and contextualize it, but the recommendation itself is always copied directly from the deterministic score.
  • Due Diligence only cites facts actually on file — it can't reference a fact_id or conflict_id that doesn't genuinely exist for that property, and a hallucinated reference is silently dropped rather than trusted. See Due Diligence.
  • Genuine AI calls take real time. Expect several seconds to tens of seconds for Ezrah, Due Diligence, or Investment Manager to respond — these aren't instant, cached, or precomputed.
  • Occasional failures happen. AI responses can, rarely, come back in a format Brickwise can't parse — you'll see an explicit error rather than a silently wrong result, and a retry typically resolves it. See FAQ.
  • No estimated dollar cost is shown anywhere. Brickwise logs real token counts for every AI call (visible in the Investment Manager panel's Agent runs table) but deliberately doesn't convert that into an estimated price, since a wrong estimate would be its own kind of fabrication.
  • Ezrah only knows what's scoped to the property or deal you're asking about. It can't answer questions about your whole portfolio or other properties from within a single property's Ask Ezrah panel.

External data providers

Market research, rental comps, sales/ARV comps, and listing-price monitoring all depend on external data providers — and in a standard configuration, most aren't connected:

  • Address geocoding works (via the U.S. Census Bureau).
  • Market data, rental comps, sales comps, and live listing-price tracking are typically not connected. These features exist and are fully built, but without a provider, you'll routinely see honest BLOCKED or INSUFFICIENT DATA messages instead of results — this is expected behavior, not a bug, and Brickwise deliberately shows you why rather than filling the gap with a plausible-looking guess.

Provenance data (Facts & Sources)

Due Diligence and the citation chips throughout the app are built on Facts tied to Sources (see the Glossary). As of this writing, there's no in-app form for adding a Fact or Source directly — a deal typically needs this data populated through another means before Due Diligence has anything to review. If a property has no facts on file, Due Diligence and the citation features will accurately show there's nothing to work with, rather than fail confusingly.

Account & security

  • Forgot your password? Select Forgot password? on the login page and follow the emailed link — see FAQ.
  • New accounts get a verification email. A banner reminds you to verify until you do; every other part of the app works normally in the meantime, verified or not.
  • Access tokens expire after 30 minutes; refresh tokens after 7 days. Brickwise refreshes your session automatically in the background — you'll only be asked to log in again after extended inactivity. Logging out revokes the session server-side, not just locally.
  • Uploaded documents are scanned for malware in the background after upload — there's no visible status in the document list, but a flagged file raises an alert on the deal.
  • Rate limiting exists on login/register/refresh but isn't visible in the UI — if requests are unusually fast, you may see a generic rate-limit error rather than the normal response.
  • You can export your data or delete your account from Settings (select your name, top right). Deletion requires your current password and can't be undone.

Data entry limits

  • Purchase price is capped at $999,999,999.99 (matching the database's storage precision).
  • Purchase date can't be in the future.
  • Most financial input fields (rates, costs, rents) have a minimum of 0. A few fields — notably the per-month rate/occupancy cells in STR Scenarios — don't currently enforce this; see Known UI Issues below.
  • Document uploads are capped at 25 MB per file.

No confirmation on some deletes

Removing a unit, a saved BRRRR/STR scenario, a due diligence task, a contact, or a document happens immediately, with no "are you sure" prompt, and can't be undone from the app. (Deleting a deal or a portfolio does ask for confirmation.) Take a moment before selecting Remove or Delete on these.

Data retention

There's no stated data-retention or auto-deletion policy — data you enter stays until you delete it yourself (where a delete option exists) or your account is removed.

Known UI issues

These are real inconsistencies or rough edges in the current interface — flagged plainly rather than described as if they were intentional:

  • The same status is sometimes styled two different ways. For example, the Investment Score's recommendation appears as friendly text ("Strong candidate") on the Scoring panel and Compare page, but as raw uppercase ("STRONG CANDIDATE") on the Investment Manager panel.
  • Document types, due diligence severities/statuses, and one of the provenance badges ("UNVERIFIED", etc.) show as raw, all-caps system values rather than the friendlier formatting used for most other statuses in the app (deal stages, contact roles, investment thesis).
  • The Compare page's Investment thesis row and its scenario toggle (Base/Downside/Severe, missing Upside) behave slightly differently from the equivalent controls elsewhere in the app.
  • Registration doesn't state the 8-character password minimum anywhere on the page — you'll only discover it if you try a shorter one.
  • The "Open underwriting" link on a deal navigates to the property's own page (not a same-page section) — expect a real page change, not an in-page jump.
  • Score ranges aren't explained on screen. The Investment Score, Data Confidence Score, and the Research panel's Income/Appreciation/Market-risk scores are all shown as bare 0–100-ish numbers with no on-screen scale or benchmark.
  • A few numeric fields lack minimum/maximum validation that similar fields elsewhere enforce — notably the per-month cells in STR Scenarios.
  • Some destructive actions skip confirmation where similar ones elsewhere ask for it — see No confirmation on some deletes above.
  • The mobile navigation menu button is icon-only (no visible "Menu" text, though it does have a label for screen readers).
  • Ask Ezrah's "Show calculations used" reveals raw JSON — genuinely useful for verification, but more technical than the rest of the app.

None of these affect the correctness of the underlying numbers — they're presentation/consistency issues, not calculation bugs.