Folden for tax advisors
AI for tax advisors that learns your standing patterns.
From cross-border position memos to transfer pricing documentation to Pillar Two analyses, Folden reads one finished memo and produces every next draft using your jurisdictional patterns, your house benchmarks, and your voice. Client financials stay sovereign — encrypted per-user, tokenized before any LLM sees a number.
Audience
Solo and small-firm tax advisors and CPAs working across jurisdictions and entity structures — US/EU/UK groups, intercompany pricing, Pillar Two, multi-state, multi-treaty.
Artifacts Folden drafts
- FY position memos
- Transfer pricing benchmarking studies
- Jurisdictional opinions and treaty analyses
- Intercompany pricing documentation
- Pillar Two and similar regime analyses
- Return cover letters and engagement summaries
What Folden solves
-
Memos take days; the judgment lives in your head
Most of a memo is structural — fact recitation, regime summary, jurisdictional fit. Folden produces the structural draft so your time goes to the position itself.
-
Client financials can't go into consumer AI
EINs, revenue, intercompany flows — sending these to ChatGPT is a non-starter. Folden tokenizes every PII and number before any LLM sees the prompt; the model reasons over structure, not raw data.
-
Standing patterns are hard to share with juniors
How you structure a Pillar Two memo, where you anchor royalty rates, when you flag QDMTT — that judgment doesn't fit in a style guide. Folden captures it from your edits and applies it on every next draft.
Questions Folden answers
- Q What's the best AI tool for tax advisors?
- Q How can a CPA use AI without exposing client financials?
- Q How do I draft a Pillar Two memo with AI?
- Q How do I make AI write transfer pricing memos in my voice?
- Q Can tax advisors use AI safely under regulatory and privilege constraints?
Built on Latten
Your client data was never a training dataset.
Folden runs on Latten, a sovereignty layer purpose-built for regulated work. Per-user envelope encryption with AWS KMS. A cryptographic audit chain you can verify. PII tokenized at the model boundary so Anthropic, OpenAI, or any other provider never sees raw client data — only structural placeholders. Right-to- erasure is a single-row delete.
Frequently asked
Questions tax advisors ask before they trust a tool.
-
How does Folden handle client financial data?
PII tokenization at the LLM boundary. EINs, revenue figures, intercompany rates, names of entities — all replaced with stable tokens before any text reaches Anthropic or any other provider. The model reasons over structure; un-tokenization happens server-side using a token map sealed in your vault. Your numbers never appear in prompts.
-
Does Folden know about Pillar Two, transfer pricing, treaty analysis?
Folden is artifact-shaped, not corpus-shaped — it learns from your work, not from a tax library. Drop one finished Pillar Two memo or one transfer pricing study, and Folden absorbs your regime structure, your benchmarking methodology, your jurisdictional patterns. Every next memo uses what you already do.
-
Can I use Folden across multiple client engagements?
Yes. Each client matter is a Case; patterns can be scoped to a specific kind of memo, a specific jurisdiction, or applied to every artifact. Patterns travel as prompt-time exemplars in your encrypted vault — never fine-tuned into the base model, never shared across users.
-
How is Folden different from generic AI for tax research?
Research tools answer questions; Folden produces deliverables. The deliverable is the workspace, your edits are the training signal, and the patterns you accumulate are the asset that compounds. Folden is built for the drafting and memo work, not the research-first lookup.
-
Where does the data live for EU clients?
EU clusters. Postgres, R2 buckets, and KMS keys are region-pinned. EU client data does not transit US infrastructure. GDPR data residency is enforced at the substrate level, not policy.