TrustAIby Bonis Systems
For solo & small trust-and-estate firms

Your firm. Your case.
Your time back.

A trust dispute lands on your desk and brings eight years of bank statements, six tax returns, three trust amendments, and a binder of correspondence with it. Your client is paying you to litigate the case, not to spend Saturday reading 1041s. TrustAI reads the documents, traces the money, drafts the petition, and tells you which of your outside experts you actually need to call. You stay on the case work.

13 analysis modes Cross-validation pass on every output Texas Trust + Property Codes + BOC embedded Vision-language reading of scans, photos, faxes Knox-anchored receipt chain
The trap a small firm walks into

Trust and estate cases punish small firms for not having a CPA in the next office.

The cases that pay best — contested trusts, fiduciary fights, multi-year estate disputes — are the same cases that drown a small firm in document review. The work that wins is buried under the work that doesn’t.

What changes when you upload a case

The 13 modes work on the same case context. You don’t re-explain the family tree thirteen times.

A case in TrustAI carries one shared context — parties, dates, accounts, trust instruments, relationships, conflicts, claims. Every analysis mode draws on that context. You upload the documents once; the modes work over them in the order you need them.

The second opinion you couldn’t afford to run on every output

Cross-validation on every output. Multi-model methodology demonstrated on the founding case.

Every analysis your firm runs gets a second independent pass — same documents, same case context, fresh review. Divergences between the two passes are flagged for human review and recorded as part of your case file, instead of disappearing into one opinion you have no way to test.

The multi-year contested trust dispute that TrustAI was built against was independently processed through three frontier models — Anthropic Claude, OpenAI ChatGPT, and xAI Grok — via their public web interfaces. The convergence and divergence among the three were ingested into the case record as cross-validation evidence. That established the methodology TrustAI is built on.

Production today runs Anthropic Claude (claude-sonnet-4-6) as primary, with OpenAI GPT-5.5 and xAI Grok running in parallel as independent cross-validators on every findings-level inference. Each vendor call emits a Knox-anchored receipt with vendor name, model, content commitment hash, and latency. Cross-checks produce a consensus anchor when shadow vendors converge with the primary, or a divergence anchor when any shadow falls below the consensus threshold. Both outcomes are publicly readable on TrustAI’s Knox chain at /api/blockchain/blocks and /api/blockchain/verify. Vision-extraction perception tasks remain Anthropic-primary because cross-vendor PDF parity is incomplete; the findings derived from extracted text run the full cross-check downstream. The shadow vendors have not been trained on the methodology this firm uses — their independent non-conformity is the basis of the cross-check, not a single-vendor self-review.

Texas-deep, 50-state extensible

Texas Trust Code, Property Code, and Business Organizations Code embedded in the analytical prompts.

If your case is in Texas, the Trust Code, Property Code, and Business Organizations Code statutory framework is in the prompts with section text. Texas Estates Code coverage is named in the prompt-instruction layer and is the next coverage build. If your case is somewhere else, the architecture is set up to extend the same pattern to your state. The prompt library is the moat; the case context is yours.

Documents in, however they show up

Scanned. Photographed. Faxed. Twice-photocopied. Read all the same.

PDF text parsing first; for scans, photographs, and faxed exhibits, every page goes to a vision-language model that reads pixel-level content and transcribes it. Every document is SHA-256 hashed on ingest so the version that drove your analysis is the version you can prove later. You don’t pre-clean documents to make the platform happy. You upload what your client gave you.

Your record outlives your subscription

Every analytical step is anchored on a hash chain that doesn’t depend on TrustAI staying online.

Every document ingested, every finding written, every cross-validation pass — appended to a SHA-256 hash chain, rolled into hourly Merkle trees, anchored to the Bitcoin blockchain via OpenTimestamps. The verification doesn’t require trusting Bonis Systems, doesn’t require TrustAI to still be online a year from now, and doesn’t require the platform you used to be the platform you go to court with. The audit trail is yours, externally verifiable, with public tooling.

The same hash chain anchors the rest of TrustAI’s parent platform. One product goes away, the chain stays verifiable. That matters when a case ages out the normal product lifecycle and the underlying record still has to hold up.

How a small firm starts

Talk to the founder directly.

TrustAI is operated by Bonis Systems LLC — an AI-native firm with principal founder Jonis Aaron Fields and an AI co-founder. There is no sales team and no procurement queue. Email or call directly; if it’s a fit, you get access to the platform and a real conversation about your case context. If it isn’t, you get an honest answer about what TrustAI does and doesn’t do today.

Pricing on request — this is a working production system, not a free trial.