B.Legal Practice Intelligence
Book a Demo
Back to Help Center Help Guide

AI & Automation

Use AI to generate billing narratives, draft documents, and research legal issues. All AI features are attorney-reviewed -- nothing is auto-saved or auto-sent.

Platform availability: AI Billing Narratives and the AI assistant run on all platforms. AI Drafting, AI Research, Email Scanning, and Duplicate Detection are available in the Mac and iOS apps. The Screen Capture feature is Mac-only (requires ScreenCaptureKit). Web parity for the rest is on the roadmap.

AI Billing Narratives

What it does

The AI billing narrative feature expands attorney shorthand into full, client-appropriate billing descriptions that read naturally on an invoice. It runs on whichever AI backend you have configured -- local Ollama on macOS by default, or Anthropic Claude as an opt-in.

How to use it

  1. Open a time entry form. Create a new entry (Cmd + N on macOS) or edit an existing entry.
  2. Type your shorthand. In the narrative field, type your abbreviated notes. Be as brief or detailed as you like. Common abbreviations the AI understands include:
    • tc -- telephone conference
    • rev -- review
    • draft -- draft / prepare
    • opp -- opposing counsel
    • MSJ -- motion for summary judgment
    • dep -- deposition
    • PL / DEF -- plaintiff / defendant
    • re -- regarding
  3. Generate the narrative. Click the AI button. The AI produces a full narrative in a few seconds.
  4. Review the result. Read the generated narrative carefully. Edit anything that does not accurately reflect the work performed.
  5. Accept or regenerate. If the narrative is satisfactory, save the entry. If not, generate again or edit the text manually.

Tips

  • The same narrative generation powers the email, Zoom, and iMessage capture pipelines, so captured suggestions arrive with a draft narrative.
  • You are always the last step: nothing is saved until you accept it.

AI Drafting

What it does

The AI drafting tool generates first drafts of legal documents based on your instructions, across 18 New York document types -- motions, demand letters, discovery requests, and correspondence -- with a refinement loop to iterate on the output. Every draft is a starting point for attorney review; nothing is filed or sent by the app.

How to use it

  1. Open AI Drafting. Open the drafting tool from the AI section of the app.
  2. Choose the document type. Select from the available NY document types -- motions, demand letters, discovery requests, client and counsel correspondence, and more.
  3. Provide instructions. Describe what the document should address: the legal arguments, facts to include, relief sought, or points to raise. The more detail you provide, the better the initial draft.
  4. Generate and refine. Generate the draft, then use the refinement loop to ask for changes until the structure is right.
  5. Review and take it to your word processor. Read the draft thoroughly, verify every citation, correct any inaccuracies, and finish the document in your word processor.

Important

Always verify case citations and legal authority in AI-generated drafts. The AI can produce plausible-sounding but incorrect citations. Cross-check every citation against Westlaw, Lexis, or the court's PACER/ECF system before filing.

AI Legal Research

What it does

The AI research tool answers legal questions in plain language with a structured, jurisdiction-aware response. It works from the model's own knowledge -- it does not search a live case-law database -- so treat it as a well-read colleague pointing you in a direction, not as a citator.

How to use it

  1. Open AI Research. Open the research tool from the AI section of the app.
  2. Select the jurisdiction. Choose the relevant jurisdiction so the response is framed accordingly.
  3. Describe the legal issue. Type a natural-language description of the issue you are researching. Examples:
    • "What is the standard for piercing the corporate veil in New York?"
    • "Statute of limitations for breach of fiduciary duty"
    • "Spoliation sanctions under FRCP Rule 37(e)"
  4. Review results. The AI returns a structured response. Any authorities it mentions are candidates for your real research, not confirmed citations.
  5. Save what's useful. Keep research results in the tool for later reference as you work the issue.

Important

AI research results are a starting point, not a substitute for proper legal research. Always verify citations, check that cases have not been overruled, and confirm that statutes are current. Use Westlaw, Lexis, or similar services for authoritative confirmation.

Screen Capture

What it does

Screen capture (macOS only) is opt-in passive time tracking. At intervals, B.Legal captures your screen and sends the capture to your configured AI backend, which infers what you were working on. The observations are merged into suggested billing sessions that you review and confirm -- nothing is billed without your sign-off.

Privacy note

Captures are screenshots, and they are analyzed by whichever AI backend you have selected. If your backend is Anthropic Claude, captures are sent to the cloud for analysis. If client confidentiality requires that screen content never leave your Mac, use the local Ollama backend -- or leave screen capture off.

How to use it

  1. Enable screen capture. Turn the feature on in Settings. Grant the macOS Screen Recording permission when prompted.
  2. Work normally. B.Legal periodically captures and analyzes your screen while the feature is on.
  3. Review suggested sessions. Open the capture review to see inferred billing sessions with a suggested matter, duration, and narrative.
  4. Confirm, edit, or discard. Confirm a session to create a time entry, adjust it first, or discard it entirely.

Tips

  • Confirm or discard sessions before quitting the app -- unconfirmed sessions are not kept permanently.
  • This feature requires the native macOS app. It is not available on iOS, the web, or PWA platforms.

Email Scanning

What it does

When Gmail is connected, B.Legal scans your mailbox and uses AI to classify emails into matter-attributed suggestions: potential time entries for correspondence you handled and task reminders for action items. Everything lands in a review queue -- nothing becomes a time entry or task until you approve it.

How to use it

  1. Connect Gmail. Go to Settings, then Integrations, then Gmail, and authorize B.Legal (see the Communications guide for detailed setup).
  2. Run a scan. Scans run on demand, and the Mac app also runs a daily scheduled scan in the evening while the app is open.
  3. Review suggestions. Each suggestion shows the email context and a proposed time entry or task, pre-matched to a matter.
  4. Approve or dismiss. Confirm the matter assignment, adjust the duration, and approve -- or dismiss suggestions that do not represent billable work.

Tips

  • B.Legal remembers which emails have already produced time entries, so the same email is never billed twice.
  • Add client email addresses to client records to improve matter matching.
  • The scheduled scan only runs while the Mac app is open -- run a manual scan after time away.

Duplicate Detection

What it does

Duplicate detection identifies time entries that may record the same work twice -- which can happen when manual entry overlaps with email, Zoom, or iMessage capture. It uses four deterministic matching strategies, and resolved duplicates are soft-deleted with an audit record. A separate merge engine handles duplicate client records.

How it works

  1. Scan for duplicates. The detector compares entries using multiple strategies to flag likely duplicates.
  2. Review flagged pairs. Compare the flagged entries and decide which to keep.
  3. Resolve. Removing a duplicate is a soft-delete with an audit record -- nothing disappears silently.

Tips

  • Duplicate detection is most useful when you use multiple capture methods. It is a backstop against double-billing.
  • For duplicate clients, use the client merge flow -- it transfers matters and repairs references across invoices and payments.

Built-in Alerts

What it does

B.Legal watches the dates that matter and tells you about them. Local notifications are scheduled for SOL dates, court dates, and task deadlines, and the app runs a set of built-in monitoring checks at launch -- SOL warnings, overdue invoices, low firm trust balance, and approaching court dates -- each firing recorded to the audit log.

How it works

  1. Deadline notifications. SOL dates, court dates, and tasks feed a local-notification ladder so warnings escalate as dates approach.
  2. Dashboard surfacing. Overdue tasks, overdue invoices, and upcoming court dates appear on the dashboard every time you open the app.
  3. Launch-time monitoring. The built-in checks run when the app launches, and each firing is written to the audit log against the matter or invoice involved.

Tips

  • Notifications are scheduled on-device -- keep the app installed on the device where you want to be warned.
  • The monitoring checks run while the app is open. For a calendar you can check from anywhere, push your court dates to Google Calendar (see the Communications guide).

AI Model Settings

What it does

B.Legal supports two AI backends: a local Ollama instance (the default on macOS) and Anthropic Claude (cloud, opt-in). On iOS the cloud backend is the default, since iPhones can't run a local model. Switching backends applies app-wide, instantly.

How to configure

  1. Open AI settings. Go to Settings, then AI.
  2. Local AI (Ollama). Install and run Ollama on your Mac; B.Legal connects to it locally. It runs offline and keeps all AI input on your machine.
  3. Cloud AI (Anthropic Claude). Select the Claude backend and enter your own Anthropic API key.

Tips

  • Local AI via Ollama is free, works offline, and keeps privileged text on your machine -- the right default for confidential work.
  • Anthropic Claude generally produces higher-quality results for complex drafting and research, but requires an internet connection and sends the request content to Anthropic, subject to Anthropic's commercial data policies.
  • Your API key is stored securely on-device in the system keychain.

Questions about AI privacy?

We take client confidentiality seriously. Read our security overview or contact us with specific questions about data handling.

Security & Privacy Back to Help Center