DONNA · Home GitHub
In thirty seconds · No video — just read

What DONNA does, in thirty seconds.

DONNA handles the work around the work. Open-source delegation orchestration for legal practice. The lawyer speaks. DONNA routes. The proof is signed. Judgment stays with the lawyer.

The thirty-second version

One spoken sentence. Five things happen. Every one is signed.

A senior partner speaks
Send Sarah the M&A precedent we used for Dubrovnik, ask her to redline by Tuesday, copy Marcus when she replies.

That one sentence is all DONNA needs. Here is what it turns into:

  1. It captures the intent. The spoken request becomes a structured Intent Decision Record (IDR): who decided, on what evidence, with what confidence — captured at the moment it happened.
  2. It routes the work. To the right person, the right system, the right tool — the precedent retrieved, the redline requested, the follow-up set, Marcus copied. The AI does the task; the lawyer keeps the judgment.
  3. It logs the time. Later: "ninety minutes on the Zenith due diligence — competition law review." DONNA captures it by voice as a Clio-ready entry, ready for month-end export.
  4. It chains the proof. Every delegated decision is signed and linked to the one before it, like pages in a notary's logbook. Tamper-evident by construction. Replayable for audit.
  5. It exports for the regulator. The whole chain replays as a structured pack — records, model identifiers, timestamps, signatures — built for regulatory review, in an open, documented format.

This is not transcription. Speech is a means, not the product. What you keep is a signed, replayable record your firm can verify — built on an open protocol, so any AI runtime can read it.

You hold the keys

DONNA is open source and self-hosted — run it against a local model and no client matter ever leaves your premises, or connect your own AI-provider accounts so your data goes only where you choose, under your own contracts. In the default cloud setup, document scanning runs on your server, and AI text is handled by US providers that don't train on your data and keep it no longer than 30 days — under GDPR Data Processing Agreements and EU Standard Contractual Clauses. See exactly what's sent — and what isn't →

Names, matters, and figures are fictional — constructed scenarios that exercise DONNA's primitives without exposing any real client material. The pattern is realistic; the data is invented.