The Adjudica User Guide

The AI does the homework. You make the decisions.

Adjudica is your firm's AI-powered case management platform for California Workers' Compensation. Upload your case documents and the AI reads them, fills your court forms, builds your case timeline, drafts your letters, and answers questions about your files — citing the exact document and page every time. This guide walks the whole platform, screen by screen. Not a "read the whole manual" person? Jump to the Quick Reference and dive in — every sidebar item has a tooltip.

What Adjudica Does

Adjudica is the case-management system of record for a California Workers' Compensation practice. The firm runs its cases on Adjudica — matters, documents, drafts, deadlines, and the AI work product all live in one workspace.

The important part: You are always in charge. The AI proposes. You review, correct, and approve. Nothing leaves the platform without an attorney's sign-off. Think of it as a very fast, very thorough first-year associate who always shows their work and never files anything on their own.

The Glass Box promise. Most AI is a black box — you put data in, something comes out, and nobody can explain what happened in between. We do the opposite. Every AI answer in Adjudica comes with a citation to the exact document and paragraph the AI relied on. You can always see why the AI said what it said, where it got the information, and how confident it is. Transparency is not a feature. It is the whole point.

Security. HIPAA-compliant, with a Business Associate Agreement (BAA) in place. All data is encrypted at rest and in transit and hosted in US-based, BAA-covered cloud data centers. Your firm's data is isolated from every other firm's — no cross-contamination, no shared models, no training on your documents, ever. And the platform never routes Protected Health Information through any AI model that isn't covered by a signed BAA.

Quick Reference

The whole platform in a dozen rows.

Read on for the details, or just start — hover over any sidebar item for a tooltip explaining what it does.

Logging In & Navigation

Logging in

Go to app.adjudica.ai and enter your email and password. Authentication is session-based, with org and team management built in.

• First time, with an invitation? Click the link in your invitation email. Your email is pre-filled — set a password and you're in.
• First person from your firm? You'll set up a firm profile, invite your team, and accept the BAA. About two minutes.
• Forgot your password? Use "Forgot Password" on the login page. The reset link expires in 24 hours — check spam before contacting support.
• Email verification. New accounts require email verification before you can access the platform.

Compliance is front-and-center: Terms, Privacy, AI Transparency, and the BAA acknowledgement live in the footer of every page.

The sidebar

The left sidebar is your home base, and it adapts to your role — it only ever shows the surfaces your permissions allow.

Collapse the sidebar with the toggle at the top when you need more room. Your profile lives at the bottom — click it to manage your name, password, security settings, and to log out.

Home — Your Ranked Caseload

Your home page isn't an alphabetical list. It's a caseload ranked by what actually needs attention today — it answers "what should I work on next?" before you ask.

• Four KPI tiles at a glance: Active Cases, Need Attention Now, Drafts Pending Review, and Settlement Pipeline value.
• The Hot List surfaces the matters that need you most, each with a plain-language reason it's ranked where it is — no mystery scores.
• Today / This Week tracks statute-of-limitations (SOL) deadlines — the red flags a Workers' Comp practice can't afford to miss.
• The Inbox routes you straight to drafts awaiting your review and to mailroom documents waiting to be classified.

The dashboard isn't trying to show you everything. It shows you what needs your attention right now. If nothing's on it, your inbox is clear.

The activity panel

The activity panel shows what's happening in the background:

• Blue — the system is working on something (OCR, an AI form-fill, building a timeline).
• Green — a task finished successfully.
• Red — something needs your attention (a document failed processing, an AI task hit an error).

Click any activity item to jump straight to the relevant page.

Matters

A matter is a case — one per injured worker, per date of injury. Open a matter and you get a single workspace for everything about it: parties, injury and insurance details, documents, drafts, the case chat, the timeline, and the knowledge graph.

Create a matter

Matters → "New Matter" → fill in the applicant name, case number, employer, date of injury, and claim type → Create.

Open a case from documents

A case can be born from a stack of paper. Drag the pleadings into a new matter and Adjudica's multimodal AI reads the actual documents — not just the filenames — extracts the parties, employer, injury, and dates, and stands up the matter pre-filled. The most tedious part of intake becomes a review-and-confirm step.

Find a matter

The matter list is searchable and filterable — by applicant name, case number, or employer; by assigned attorney; by status (Intake, Active, Pending Settlement, Settled, Archived); and by date range. Sort any column by clicking its header.

Matter Home

Click into a matter and you land on Matter Home — everything about the case in one view. Summary cards across the top show Documents, Drafts, Events, and Days Active; click any to jump to that section. Quick actions put Upload Document and Create Draft one click away. Case details show the applicant, employer, insurance carrier, assigned attorney, involved parties, and current status.

Insurance & claim details. Each matter tracks the workers'-comp carrier, claim number, and adjuster — pulled from the case documents, and brought over from MerusCase during onboarding where available. Role-based capabilities decide what each team member can see and do; the interface never offers an action the server will refuse.

Case status lifecycle

Migrating from MerusCase

Moving onto Adjudica from MerusCase? The integration is a one-way migration bridge: during onboarding, connect your MerusCase account (Settings → Integrations) and Adjudica imports your existing cases and their documents. On any matter you can hit "Import from MerusCase" to pull case details and files in. The goal is simple — the practice runs on Adjudica, not alongside a legacy tool.

Mailroom

Every document starts here. The Mailroom is a transparency-first intake pipeline that shows exactly where each file is and why — the heart of the Glass Box.

Upload

1. Go to Mailroom.
2. Drag and drop files into the upload area (or click to browse).
3. Assign each document to a matter — start typing the applicant name, case number, or employer and pick from the dropdown.
4. Click Upload.

Supported formats: PDF, DOCX, JPG, PNG, and TIFF — up to 50 MB per file. PDFs give the best results by a wide margin.

Three lanes, so the pipeline is legible

Documents move through stages automatically. Three lanes make the whole pipeline readable at a glance:

A file that can be re-processed shows a Retry; one that genuinely can't (encrypted or unsupported) doesn't pretend it can. Filter by document type, subtype, or matter, search by filename, and select multiple documents to classify or route them in bulk.

Tips for better results

• Better scans = better AI. If your fax from 2003 looks like it went through a sandstorm, the AI will tell you it can't read it. Honestly, neither can you.
• Upload everything relevant to a matter before running AI features — the more the AI has to work with, the better it performs.
• Attorney-client privileged material? Only upload it if you want the AI to read it. The AI doesn't know something is privileged; it reads and cites it like any other document.
• You can always add more documents later. New uploads become available to AI features as soon as they're processed.

How it works (the Glass Box). When you upload a document, the system first converts it to searchable text with optical character recognition (OCR) — teaching the computer to read the page the way you would — then corrects the result with an AI model that identifies the document type (QME report, medical record, hearing notice) and tags it. Clean scans read in seconds; a page that's been photocopied three times gets lower confidence scores. Every classification carries a confidence score, and nothing is finalized until a human reviews and approves it. What this means for you: you upload, the AI reads and sorts, you approve. The heavy lifting is done, but the final call is yours.

Documents

Documents shows every case file across every matter in the firm — a single, searchable system of record.

Browse and find

Search by filename; filter by matter, document type, subtype, status, or date range; paginate at 25, 50, or 100 per page. When you're hunting a specific QME report across the whole caseload, this is where you go.

The document viewer

Open any document and Adjudica shows it alongside what the AI understood about it — type, classification, and the extracted evidence — never a black box. The original file sits next to its AI-derived metadata, so you can verify the machine's read, and each document carries an explainer: why was this classified this way? PHI-sensitive fields (AI reasoning, extracted values) are access-controlled on the server, not just hidden in the interface.

OCR review — the Glass Box gate

Before any AI-extracted text is trusted, an attorney reviews and corrects it side-by-side with the source page. The system displays the extracted text paragraph by paragraph, each with a confidence score.

How it works (the Glass Box). Each paragraph gets a confidence score based on how clearly the AI could read the text. Very High (80%+) means the text was clean. Low (below 30%) means the AI is squinting — handwritten notes on a prescription pad, faded carbon copies from the 1990s. The score measures legibility, not legal accuracy: the AI is telling you how well it could read the page, not whether the content is legally correct. Low-confidence paragraphs are flagged so you know where to focus. What this means for you: skim the high-confidence paragraphs; spend your time where the AI struggled — those are the ones most likely to contain errors.

Approve accurate extractions, correct mistakes, or delete content the AI misread. Your approved version becomes the official extracted text the AI uses for form filling, timeline building, and chat. This step matters: the quality of every downstream AI feature depends on it. Five minutes correcting a low-confidence extraction now saves you tracking down wrong answers later.

Drafts

This is where documents get made — the core of the platform. Three paths, one goal: get your documents done faster without sacrificing accuracy.

To create any draft: matter → Drafts → "Create Draft" → pick your type.

Court Form drafts

The feature that saves the most time.

1. Open your matter → Drafts → "Create Draft".
2. Select "Court Form" and pick a DWC form from the library.
3. Click "AI Fill".
4. Wait two to four minutes (longer for matters with many documents).
5. Review the completed form — every field is filled in or flagged.

Confidence indicators. This is how the AI tells you how sure it is. Your job is to verify.

Green does not mean correct. It means the AI thinks it's correct. The AI is confident, not infallible — much like that one opposing counsel we all know.

Multi-candidate answers. Sometimes the AI finds more than one possible answer for a field. You get a dropdown of candidates, each showing which document and paragraph it came from. Pick the right one, or type your own.

Source citations. Every AI-proposed answer links to the exact document and paragraph it relied on. Click the citation to see the source text — the receipt. If the receipt doesn't support the answer, override it. That's the entire review workflow: if an answer looks suspicious, click the citation; if the source says what the AI claims, move on; if not, fix it.

Override any field. Click the edit icon, type your correction, done. Your change is logged in the audit trail — there's always a record that you reviewed and modified the AI's output.

How it works (the Glass Box). When you click "AI Fill," Adjudica reads every document in your matter and, for each field, searches for the specific passages that answer it, evaluates the candidates, proposes the best-supported answer, and scores its own confidence — green when the evidence is strong and consistent, yellow when the support is thinner, red when it's essentially guessing, blank when it found nothing. Form-filling runs on Google's Gemini models; OpenAI is never used anywhere claimant PHI is processed, because it isn't covered by a Business Associate Agreement. The AI only reads documents in your matter — never other firms' data, external databases, or anything you haven't uploaded — and every proposed answer links to the exact paragraph it came from, with an "AI-generated — verify before use" disclaimer. What this means for you: you get a first draft in minutes instead of hours. Your job is to check the AI's homework, not do it from scratch.

Template drafts

Select a firm template — one you've created or one from the pre-built library. The AI reads the matter's case data and documents, fills the placeholders (applicant name, date of injury, employer, medical providers), and presents the draft for review. Great for demand letters, MSC statements, and anything the firm writes repeatedly. The AI does a good job with factual placeholders; narrative sections may need more of your attention — it writes competently but generically, and your voice and strategy make the difference.

Custom drafts

Describe what you need in plain language — for example, "Write a letter to the carrier requesting authorization for the recommended surgery based on the QME report." The AI reads your case documents, finds the relevant information, and generates a complete document grounded in your uploaded files. Best for one-off documents where no template exists yet. If you find yourself creating the same custom draft repeatedly, that's a sign to turn it into a template.

The editor

All three draft types open in the same editor:

• Auto-save — your work is saved continuously.
• Refine with AI — select any block of text and rewrite just that portion: tighten the phrasing, adjust the tone.
• Preview PDF — see exactly how the document will look before exporting.
• Export to PDF or Word (DOCX).
• Publish — finalize the draft. Published drafts are locked and versioned; create a new version if you need changes later.

A note on rendering. Always download and review the final PDF before filing. Formatting occasionally shifts between the editor and the exported PDF — a table cell wraps differently, a page break lands awkwardly. Check the PDF. Every time.

Court filings

When a document is ready, Adjudica assembles the filing package, binds the attorney's certification to its exact contents (a SHA-256 hash of exactly what's certified), and prepares it for submission. Certification is an explicit attorney step — identity bound to the reviewed bytes — so a human stays accountable for what's filed. Electronic filing is in preview today, with a court-filing integration in progress; the surface is built and ready for it.

Pre-filing checklist

The AI did the first pass. Before filing or serving anything:

• Every field reviewed against the cited source document
• Every red or blank field completed with verified information
• All dates accurate and in the correct format
• Dropdowns and radio buttons reflect the legally correct option
• Generated PDF downloaded and reviewed as a complete document
• A licensed California attorney has reviewed and approved the form
• The form reflects the most current case information

The AI shows its work. Your job is to check it. That's the deal.

Timeline

Upload your case documents and the AI builds a chronological timeline — no manual data entry. It reads everything you uploaded, finds dates paired with events, and plots them in order with descriptions, categories, and citations back to the source.

Event types

Eight color-coded categories let you scan the history at a glance:

How to use it

Go to Timeline, select your matter (type to search), and events appear as documents are processed — the more you upload, the more complete the timeline. Filter by category or date range, use the zoom controls to focus on a period, and click Jump to Present for the most recent events.

Duplicate merging. If three documents mention the same hearing on the same date, you get one entry with three citations — not three duplicates.

Details on click. Click any event for the full description, all source citations, and links to the original documents. Click a citation to jump to the exact paragraph.

Add or edit manually. The AI will miss things not in your documents — phone calls, verbal agreements, things you just know. Click "Add Event" to create one, or edit any existing event.

AI chat on the timeline. Click the sparkles icon while viewing the timeline and ask questions like "Summarize the treatment history" or "What happened between the QME evaluation and the hearing?" — great for MSC prep and getting up to speed on an inherited case.

How it works (the Glass Box). The AI reads every document in your matter, looks for dates paired with descriptions of events, categorizes them into the eight types, and plots them chronologically — merging multiple mentions of the same event into one entry with citations to each source. The timeline is only as complete as your uploaded documents; if a key document is missing, the event won't appear. What this means for you: upload everything first, then build the timeline.

Case Chat

Ask questions about your case in plain English. The AI searches all your uploaded documents and answers with citations — like a research clerk who has read every page of every file and can find anything in seconds.

Click the sparkles icon on any matter page, type your question, and get an answer with links to the exact paragraphs the AI relied on.

Good questions

• "What does Dr. Johnson's QME report say about work restrictions?"
• "Summarize the treating physician notes from January through June."
• "Which document mentions the average weekly wage?"

Not-so-good questions

• "Should I take this settlement offer?" — the AI doesn't give legal advice. That's your job, and it always will be.
• "What does Labor Code 4660.1 say?" — case chat only reads your uploaded documents, not external legal databases. (For research across CA WC law, see Ask Adjudica.)
• "What did the client say on the phone last week?" — if it isn't in an uploaded document, the AI doesn't know. It reads files, not minds.

Draft chat. While editing a draft, the sparkles icon opens a chat scoped to that specific draft — ask about form fields, request language for narrative sections, or ask the AI to explain why it chose an answer.

A few details worth knowing: each chat session starts fresh (the AI doesn't carry context between conversations — a deliberate choice that prevents it from carrying forward misunderstandings); always verify citations; and be specific — "What body parts were injured according to the QME report dated March 15, 2025?" gets a far more useful answer than "Tell me about the injury."

How it works (the Glass Box). When you ask a question, the system searches your case documents for the most relevant passages — not by keyword matching, but by understanding what you mean — reads the top results, and composes an answer that cites exactly where it found the information. It only accesses documents within your matter, and each session is independent. What this means for you: fast case research with receipts. Ask, get an answer, click the citation to verify. Repeat.

Case Intelligence & the Knowledge Graph

As documents flow through the pipeline, the AI does more than read and classify — it builds a knowledge graph: a structured map of every person, organization, body part, claim, treatment, and relationship it finds across your case files. Case Intelligence is where you see that map, verify it, and act on it. The knowledge graph is rolling out now; where it's active, its entities and edges populate from the live case record.

Open it from the Case Intelligence button in the matter header — a panel slides in with four tabs.

Overview

Your at-a-glance intelligence dashboard: a maturity bar (Nascent → Developing → Established → Mature) showing how much the AI has learned; a coverage metric for how many documents have been analyzed; summary cards for entities found, relationships mapped, and attention items; an entity breakdown by type; and View Full Graph to open the interactive explorer.

Entities

Browse every entity the AI extracted, searchable and filterable by type, confidence, or source document. Each card shows the name and role, a confidence badge, document and alias counts, and a contradiction indicator.

Click any entity to expand and see all properties, clickable source documents, aliases, and related entities.

Relationships

Browse the connections between entities — who treats whom, who employs whom, which body part was injured in which incident. Each card shows the connection (e.g. "Dr. Garcia TREATS John Smith"), a weight bar for how strong the connection is, a corroboration count, a traversal count (how often the AI has used the relationship to answer questions), verified/locked indicators, and contradiction flags with a Resolve button.

Attention Required

Your action queue, with a badge count so you always know if something needs review. Three kinds of items:

• Date conflicts — two documents disagree about a date. See both sources side by side and choose: Keep A, Keep B, Keep Both, or Dismiss.
• Merge candidates — the AI thinks two entities might be the same ("Dr. M. Garcia" and "Dr. Maria Garcia"). Merge, mark as different, or review later.
• Low-coverage areas — the AI hasn't found entity types it would expect given the case size, hinting at missing documents.

Every action here is recorded in the audit trail, and resolving a contradiction changes how the AI answers questions about that topic going forward.

How it works (the Glass Box). The graph is built automatically during document processing: as each document is read, the AI extracts entities and the relationships between them, merges repeated mentions, strengthens its confidence when sources agree, and flags conflicts for your review instead of guessing. The graph also learns from use — every time a relationship is queried it grows slightly stronger, and relationships never queried gradually weaken, so the map adapts to how you actually work the case. What this means for you: the AI builds the map; you verify the territory. Every entity, relationship, and conflict resolution is your call.

Graph Explorer

View Full Graph opens an interactive visualization of the entire case. Nodes are entities (size = how many documents mention them, color = type, transparency = confidence); edges are relationships (thickness = strength, color = status — green verified, red contradiction, gold locked). Drag to rearrange, scroll to zoom, filter by type or status, and right-click a node to Verify, View Sources, or Merge. It's the same data as the panel tabs, visualized — useful for seeing the shape of a complex case and where the gaps are.

Routing transparency

When the AI uses the graph to answer a question, you may see a "how this answer was found" indicator. Expand it to see the route (did it hit the knowledge graph, fall back to document search, or combine both?), the tier (an instant lookup, a single-hop traversal, or multi-hop reasoning), the confidence, and the full pipeline steps. This is the Glass Box at work — not just the answer, but exactly how it was found.

Templates

Build reusable documents — demand letters, settlement proposals, client correspondence — with smart placeholders that auto-fill from case data. Build once, use on every case.

Create a template. Templates → "Create New Template", then choose a starting point:

• From Samples — upload your best demand letter; the AI converts it into a reusable template with placeholders where case-specific information goes.
• From Scratch — describe the document type; the AI generates a starting structure you refine.
• From an Existing Document — upload any document and the AI suggests where placeholders should go (applicant name here, DOI there); you approve or adjust.

Edit and organize. Use the editor to add, remove, or reorder sections; drag-and-drop to rearrange; preview with real case data; save. Placeholders like [Law Firm Name] and [Date] mark the raw inputs the AI resolves per matter.

Pre-built templates. A starter set ships for common WC documents — demand letters, MSC statements, C&R cover letters, client correspondence, discovery requests and responses, petitions, subpoena cover letters. Clone any of them and make it yours; the originals stay untouched.

Use a template. Matter → "Create Draft" → "From Template" → pick the template → the AI fills the placeholders → review and adjust → export as PDF or Word.

Workflows

Set up rules that do things automatically — the kind of thing a really organized paralegal would do, except it happens instantly every time and never forgets. The platform states each rule in plain English: when a case condition is met, generate a draft — and route it for human approval.

Example: "When a new medical report uploads to any matter, automatically generate a Medical Summary draft." The draft appears, pre-filled from the case documents, ready for review. No manual step to kick it off.

Set one up. Workflows → "Create Workflow" → pick the trigger (the document type/subtype that fires the rule) → pick the action (the template or form to generate) → Enable and Save. A workflow needs at least one template to fire, which is why Templates and Workflows are designed as a pair.

The Common Workflow Library. A catalog of pre-built California WC automations you can deploy in one click — each a ready-made rule (e.g. auto-generate a Medical Summary when a medical report is approved). Only admins can deploy or retract them.

Manage. Toggle any workflow on or off; the page shows the last-fired timestamp for each rule; edit or delete as your processes change. And remember — the workflow creates the draft, but a human still reviews it. Automation handles the initiation; quality control is still yours.

Productivity & Coordination

Beyond the case work, a set of capability-gated tools help you run the day. Each lives under Tools in the sidebar and appears only for roles whose permissions allow it.

• Worklist — a consolidated list of what's on your plate across the whole caseload. Home tells you which cases need attention; the Worklist tells you the specific tasks, so nothing slips.
• Calendar — the time view of the practice: appointments, hearings, and the statutory deadlines that drive a WC matter, in a date-oriented layout. It surfaces the same SOL deadlines that flag on Home.
• Reports — firm-level analytics that aggregate caseload, pipeline, and throughput into reportable views — the metrics a managing attorney uses to see how the practice is performing.
• Contacts — a firm-wide directory of the people and organizations a WC practice deals with, in one searchable place.
• Vendors — a management surface for the copy services, providers, and court e-filing partners a firm coordinates with. Scoped vendor access ships as a visible preview, deliberately inert until a properly scoped, time-boxed, audited access model is in place — least-privilege by design, with honest labeling throughout.

Administration

Firm administration is visible only to users with the Admin role. If you don't see Settings in the sidebar, you don't have Admin access — talk to whoever set up your firm's account.

Firm details. Edit the firm's name, address, phone, and contact information — this appears in templates and generated documents that reference firm details.

Team members. Invite people by email and assign a role; resend invitations to anyone who hasn't accepted; remove members who've left. Each user gets exactly one role.

Roles. Capabilities are membership-driven and enforced across the whole app.

Role Management gives a read-only map of roles → permissions → members, so who-can-do-what is legible at a glance.

Matter Deny-List. Some cases need a smaller circle. Settings → Matter Deny-List → pick the matter → select which members should not see it. Those users won't see the matter anywhere — not in the list, search, or dashboard — and it controls the AI's view of the matter too, a real data-isolation boundary. Reversible at any time.

Case Origination config. Tune how new business is scored and intaken — scoring-pillar weights, retainer/referral templates, and DWC-1 intake mode — the engine behind conflict checks, document generation, and case scoring for prospective matters.

Integrations. Where external systems plug in. MerusCase import is first-class: connect via OAuth (per-firm credentials, admin-gated, tokens stored encrypted) and migrate existing cases and claimant documents onto Adjudica. The document/OCR pipeline configuration lives here too.

Your Profile & Security

Click your name at the bottom of the sidebar to reach your profile.

Personal information. Update your name, phone, California Bar number, and profile photo. Your Bar number flows into generated documents and filings — keep it current. Your email is read-only; it's your account identity.

Password. Change it any time; you'll enter your current password first, then the new one twice. Use something strong and unique.

Two-factor authentication (2FA). Two-factor is like a deadbolt on your office door — someone needs your password and your phone to get in.

1. Go to Profile → Security.
2. Click "Enable Two-Factor Authentication."
3. Scan the QR code with an authenticator app (Google Authenticator, Authy, Microsoft Authenticator, or any TOTP-compatible app).
4. Enter the 6-digit code to verify.
5. Save your backup codes somewhere safe — a password manager, or a printed sheet in a locked drawer. They're your emergency key if you lose your authenticator.

After enabling 2FA you'll enter a code at each login in addition to your password. We strongly recommend it for every account that handles client data — it takes thirty seconds and significantly reduces the risk of unauthorized access. For the most privileged roles, it's required.

The Adjudica Ecosystem

Two companion tools sit alongside the case-management app:

• Ask Adjudica — an AI legal-research surface tuned for California Workers' Compensation. Ask a question and get an explainable, sourced answer rather than a confident guess. It runs on multi-model orchestration across Anthropic Claude and Google Gemini — the right model for each job, and never OpenAI in any PHI path. Answers are built to be traceable back to their sources, in keeping with the platform's transparency standard, and it's the foundation for deeper research capabilities on the way.
• The PD Calculator — a free permanent-disability rating calculator for California WC, open to anyone.

Both reflect the same principle as the rest of the platform: show your work.

When to Override the AI

The AI is a tool, not an authority. Always override its suggestions when:

1. You have information the AI doesn't — verbal conversations, privileged strategy, settlement terms discussed off the record. The AI only knows what's in the uploaded documents.
2. The case has unusual facts — atypical injury mechanisms, complex apportionment, novel legal arguments. The AI looks for patterns; unusual cases can produce answers that are technically sourced but strategically wrong.
3. The citation doesn't support the answer — click it, read the source paragraph, and if it doesn't say what the AI thinks it says, override the field.
4. Legal strategy requires framing facts differently — the AI presents facts neutrally; your job often requires framing them persuasively. An AI-generated description is a starting point, not final copy.
5. In doubt — your professional judgment always wins. The AI doesn't have a law license. You do.

The pattern is simple: trust the AI to find information, but trust yourself to interpret it. The AI is an extraordinarily fast reader with no legal judgment; you're a slower reader with decades of training in exactly the judgment that matters. Together, you're faster and more accurate than either of you alone.

Support

Hours: Monday–Friday, 9 AM to 5 PM Pacific. Security is monitored 24/7.

Reporting an AI error. If the AI produces an answer that's clearly wrong — not just low-confidence, but factually incorrect given the source document — email support@adjudica.ai with the subject "AI Error Report" and include the matter name or ID, the field or question where it happened, what the AI said versus what the source actually says, and a screenshot if possible. Every report is reviewed and directly improves the platform.

The AI does the homework. You make the decisions. That is how it should be.