Skip to main content
Wealth Recon
Apply for access

How Wealth Recon works

Three steps. Fifteen minutes. One dossier.

Step 1: Submit a name and a place of work

You enter the prospect's first name, last name, state of residence, and at least one of (current place of work, previous place of work). Optional fields help the engine narrow down to the right person: employer plus title, alma mater, LinkedIn Uniform Resource Locator, likely nicknames, additional context.

The form's framing copy reads, prominently, above the optional-fields section: "The more you tell us, the better we can find the right person. Optional fields help us filter out the noise of other candidates with similar names. Without your help narrowing down, the system may return a dossier with incomplete information, or worse, on the wrong person. If you are unsure about a field, leave it blank rather than guess."

The disambiguation pre-pass runs before any credit consumes. The pass returns one of four outcomes:

  • Single high-confidence match. Your dossier enters the queue. One credit consumes.
  • Single match, dossier already in your firm under twenty-four hours old. The existing dossier opens; no credit consumes. You can request a fresh rebuild for one credit if the existing dossier needs an update.
  • Multiple plausible matches. Up to six candidate cards render with name, employer, city, approximate age, and LinkedIn handle. Pick the right candidate; one credit consumes. Or return to the form and add the optional fields that would narrow this down. No credit consumes until you select.
  • Zero plausible matches. No credit consumes. The form returns with a candor note explaining what we could not determine and which optional fields would most likely resolve the gap.

You never spend a credit on a dossier we cannot start.

Step 2: Wait fifteen minutes

The pipeline runs eight artificial-intelligence agents in coordinated sequence on the V2 hybrid quality stack:

  1. Disambiguation pre-pass on Google Gemini 2.5 Flash. Resolves your input to a single living adult.
  2. Research agent on Perplexity Sonar Pro. Pulls source material across Securities and Exchange Commission filings, court records, real estate records, Federal Election Commission contributions, Internal Revenue Service Form 990 plus 990-PF nonprofit filings, news, professional history, and the rest of the public-source corpus.
  3. Drafting agent on Anthropic Claude Opus 4.7. Turns the verified research bundle into the prose an advisor reads, structured into the canonical fifteen-section dossier.
  4. Review agent on Perplexity Sonar Pro. Re-fetches every cited Uniform Resource Locator and re-confirms that the source actually supports the claim. Different model from the drafting agent so a fabrication does not get caught by the lab that produced it.
  5. Adversarial review agent on xAI Grok-4. Hunts for unsupported claims, weak citations, contradictions, and load-bearing inferences with shaky source backing. Different lab from the first two so the adversarial pass catches what they missed.
  6. Visual quality assurance on Google Gemini 2.5 Flash and 2.5 Pro. Cascading two-stage check on the rendered HyperText Markup Language and downloaded Portable Document Format.
  7. Confidence-scoring agent on Anthropic Claude Haiku 4.5. Computes a four-signal composite from coverage, citation density, identification certainty, and net-worth-band tightness.
  8. Remediation agent on Anthropic Claude Haiku 4.5. Watches every other agent and intervenes on infrastructure errors, application-programming-interface timeouts, source provider outages, and stalls.

Five different artificial-intelligence labs touch every dossier (Anthropic, Google, Perplexity, xAI, plus the Securities and Exchange Commission's public Form ADV bulk download). Every claim survives the multi-model gauntlet before it ships.

The dossier ships when every claim carries a verified source. If the confidence score lands below eighty, the dossier still ships, the credit refunds automatically, and a Low Signal banner names the gap.

Step 3: Walk in knowing

The dossier opens in the in-app reader. Two views:

  • Research view is the default canonical surface. Facts plus citations, edit affordances, source manifest, audit log access, full claim-level interaction. This is where the advisor prepares.
  • Advisor view surfaces four "ask-about" talking points the drafting agent emits alongside the structured fields. The household sketch and the at-a-glance line stay; the rest collapses to the prep an advisor needs in the back of an Uber on the way to the meeting.

You can edit any fact inline. Edits save continuously. Your name, your timestamp, attached to the edited element. Your edit informs every future dossier in your firm on the same prospect.

When you are ready, download the styled Portable Document Format with one click. Or export the compliance-ready Comma-Separated Values file: every claim, every Uniform Resource Locator, every verification timestamp, every editor attribution, in eight columns ready for a compliance officer's review.

The Source Manifest at the back of every dossier is a flat ledger. No narrative. No editorial framing. Each section heading mirrors the corresponding dossier section title; below each heading the manifest lists every Uniform Resource Locator cited in that section. Click any Uniform Resource Locator to open the source in a new tab and verify the claim yourself.

Walk in knowing.

[CTA Button: Apply for access]

What happens between dossiers

Wealth Recon caches every dossier we produce. Three rules govern how the cache behaves:

  • Under twenty-four hours. The cached dossier returns silently. No agents run. No credit consumes for a same-firm advisor opening the dossier; the original requesting advisor's credit was already consumed at the original request.
  • Twenty-four hours to thirty days. The refresh agent runs in the background on Anthropic Claude Sonnet 4.6. Time-sensitive sources (news, Securities and Exchange Commission filings, court records, Federal Election Commission contributions, recent real estate transactions) re-check; stable sources carry forward. The freshened dossier returns inside the five-minute incremental update target.
  • Older than thirty days. Full pipeline runs from scratch.

You see only your own request date on every dossier you open. No advisor learns that another advisor pulled the same prospect, including same-firm. The cache reuse is disclosed in the Terms of Service.

What we do not do

We do not buy data from breach markets. We do not pay third parties operating outside the law. We do not access medical records, beneficial-ownership reports filed with the Financial Crimes Enforcement Network, or social media of minor children.

We do not sell your dossier data, your edits, your advisor profile, or any other information you submit. The only people who see your dossiers are advisors at your firm. The only people who see your edits are advisors at your firm. The only person who sees aggregate cost data and platform metrics is Bryce.

We do not promise a dossier on every name. Some prospects have thin public footprints. Some are non-United States citizens we do not yet cover. Some are sealed estates or dynasty trusts that hide everything we would want to surface. When we cannot produce a useful dossier, we ship a Low Signal dossier with the gap visible, refund the credit, and let you decide whether to push further.

Why this is better than doing it yourself in Claude, ChatGPT, or Perplexity

The honest comparison matters. A capable advisor can absolutely produce a first draft on a general artificial-intelligence chat tool. We even publish a public starter prompt that gets you a respectable sixty to eighty percent of the way there.

Wealth Recon wins on the parts the prompt cannot do for you:

  • Disambiguation runs before research consumes a credit. A general chat tool will quietly blend two people with the same name into one dossier. We resolve to a single living adult or refund.
  • Every claim ships with a verifiable public source. A general chat tool will generate confident sentences without sources, or with sources that look real but were never visited. Our review agent re-fetches every cited link and confirms that the source actually supports the claim.
  • Five independent labs review every dossier. Anthropic Claude Opus 4.7 drafts, Perplexity Sonar Pro verifies, xAI Grok-4 attacks, Google Gemini 2.5 inspects. A single-model conversation cannot do that.
  • Saved history, refresh logic, exports. Your dossiers live in a tenant-safe library, refresh silently when the underlying world changes, and export as compliance-ready Portable Document Format and Comma-Separated Values. A chat tab forgets you the moment you close it.
  • Cost controls. Per-dossier, per-month, and overage caps enforced at the credit level. A general chat tool will happily burn tokens on the wrong person.
  • Advisor corrections persist. Your edit, your name, your timestamp, propagated to every future dossier on the same person across your firm. A chat tool forgets your correction when the conversation ends.
  • Audit-ready manifest. A flat ledger of every claim, every link, every verification timestamp. A compliance officer can read it. A chat transcript cannot.

If you pull one prospect in your life, use a general chat tool. If you pull a prospect every week, use Wealth Recon.

[CTA Button: Apply for access]

End of how-it-works copy.