Your AI handles matter admin

so you bill more hours.

Confidentiality, privilege, and why most AI tools don't qualify

The duty of confidentiality under the Legal Profession Uniform Law is non-negotiable, and client legal privilege survives whatever software a firm uses. The problem with generic AI tools is that they treat client data the way a search engine treats a search query — sent to a third-party model, retained for training, and processed in jurisdictions that don't recognise privilege under Australian law. SydClaw is built differently. Every piece of client data is tokenised before it reaches an AI model: names, ABNs, TFNs, addresses, and matter references are replaced with reversible tokens at the privacy router layer. The model never sees real PII. The reverse-mapping happens only inside the firm's own dedicated infrastructure.

For matter-related correspondence, this means the model can summarise a thread, draft a response, or extract key dates without any external service ever seeing your client's actual name or contact details. The output you read is detokenised back to plain text only inside your browser session. For a partner who has spent twenty years protecting privilege, this is the difference between a tool you can use and a tool you can't.

Conflict checking, intake, and the boring work that loses you matters

Most firms lose matters not because of fees but because intake takes too long. By the time the conflict check is done, the precedent is pulled, and the engagement letter is out, the prospect has called the next firm on the list. SydClaw runs intake in parallel: it cross-references the new matter against every existing matter, every prior client, and every party-of-interest the firm has touched, in seconds. It drafts the engagement letter from your firm's template, populates the costs disclosure, and routes the matter to the right partner based on practice-group rules.

For litigation files, the document and email modules ingest everything the client sends — discovery, correspondence, pleadings — and build a chronology automatically. When a junior needs to brief counsel three weeks before trial, the chronology is already there, indexed, and searchable. The AI handles the work nobody bills for, so the work that does bill happens faster.

Trust accounts and what we won't do

We do not touch trust accounts. We do not draft trust account ledger entries, we do not initiate trust transfers, and we do not propose distributions. Trust account regulation under Legal Profession Uniform General Rule 2015 is too unforgiving for an AI to live near, and the reputational risk to the firm is asymmetric. What we do is read the trust ledger to flag mismatches against client matters, draft client statements when authorised, and pre-fill the cost disclosure forms that have to go out alongside trust receipts. Anything that moves money goes through a partner with two-factor verification — not the AI.

Time recording, billing, and the AFR's billable hour

The billable hour isn't going anywhere, regardless of what the AFR predicts each year. SydClaw's time recording layer captures billable activity passively: every email drafted, every document generated, every call summarised is logged against a matter with a defensible time entry. The lawyer reviews the daily timesheet, makes adjustments, and submits. We've seen firms recover 25–40% on previously-unbilled work — the small five-minute tasks that disappear into the gaps between proper time entries.

For pricing, we charge $360 per user per month plus a $5,000 setup fee. A typical 10–20 person firm runs $5,000–$15,000 per month all-in. We do not take a percentage of recoveries, we do not charge by matter, and we do not charge for the work the AI does — only for the seats accessing it.