APEX schedules hearings, drafts decisions, and runs case research — by conversation. Built for administrative law judges, case managers, and intake staff. Thirty specialized AI tools. Every action audited. Every citation grounded.
Checking Judge Patel's calendar over the next 14 days. Predicting hearing duration from comparable cases…
| # | Date | Time | Room | Score |
|---|---|---|---|---|
| 1 | Tue May 26 | 09:00–10:30 | 3A | 0.92 |
| 2 | Wed May 27 | 14:00–15:30 | 3A | 0.88 |
| 3 | Thu May 28 | 09:00–10:30 | 2B | 0.85 |
Ask APEX to schedule, search, draft, or research — in plain English. It chooses the right tool, fills in the gaps, and proposes the action before doing it.
Every AI invocation is logged with the calling user, the tool, the parameters, and the outcome. Citations are grounded in real knowledge-base entries. Fabricated citations are forbidden by policy and verified by the system.
Mutations require user confirmation. Tools are filtered by role. The model knows what it doesn't know — and says so plainly.
APEX doesn't dump a chatbot in front of your existing system. It orchestrates the intelligent agents that already live in your case-management stack — scheduling, ALJ assignment, decision drafting — and routes through them with full audit trail.
Open a case. Type a question or an intent. No forms. No menus.
› what cases is Judge Blanda working?
The model maps your intent to the right capability. Smart search, slot finder, decision drafter, cascade impact analyzer — thirty in all.
› smart_search → name-aware ALJ filter
For anything mutating, APEX proposes — never decides. You see the parameters, the confidence score, the chosen ALJ. Click confirm or cancel.
› ⚠ Action card → ✓ confirmed → logged forever
Each tool is purpose-built, role-scoped, and audited. Selected highlights below — the full catalog ships in the demo.
Score-ranked slots across ALJ availability, courtroom conflicts, interpreter needs, and hearing-format constraints. Predicts duration from your historical record.
Ranks judges by workload, jurisdiction, specialization, and COI. Surfaces a justification with every recommendation.
Generates a realistic mock transcript grounded in the case facts — for counsel prep or judicial rehearsal. Stays inside the record. Never invents testimony.
Drafts decisions from case context, prior filings, and applicable precedent. Every citation traces to a real entry. Every claim shows its source.
Synthesizes legal memos from the federal knowledge base. Vector + symbolic search. Refuses to answer when the corpus is silent.
Before you reschedule a hearing, APEX shows every downstream hearing it would affect, every conflict it would create, and every buffer it would violate.
Live fairness score across your bench. Spot overloaded judges before the next assignment.
Long-term observations the AI remembers between conversations — pinnable, dismissable, audited.
One toggle disables every AI capability across the system. For incidents, audits, or maintenance windows.
Four shots from the running system. Each one is the actual UI — not a mockup, not a marketing render. The case data is from a live demo deployment.
smart_search,
runs it with the name-aware ALJ filter, and returns the workload summary with
live citations. Right inspector shows what sources were consulted.
APEX writes the audit log of record. Every AI action, every parameter, every outcome — captured before the action even runs. Independently queryable. Exportable.
| created_at | agent | action | entity | status |
|---|---|---|---|---|
| 2026-05-20 09:14:02 | j.patel@oah | SCHEDULE_HEARING | Hearing | OK |
| 2026-05-20 09:13:48 | j.patel@oah | FIND_AVAILABLE_SLOTS | Case | OK |
| 2026-05-20 09:13:12 | m.kim@oah | GENERATE_DECISION_DRAFT | Decision | OK |
| 2026-05-20 09:12:55 | s.rivera@oah | RESCHEDULE_HEARING | Hearing | OK |
| 2026-05-20 09:11:30 | l.chen@oah | CLASSIFY_DOCUMENT | Document | OK |
| 2026-05-20 09:10:18 | j.doe@oah | CANCEL_HEARING | Hearing | DENIED |
| 2026-05-20 09:09:55 | m.kim@oah | WRITE_CASE_MEMORY | CaseMemory | OK |
APEX does not replace decision-making. It removes the friction between you and the work that only you can do. When the model isn't certain, it asks. When the corpus doesn't support an answer, it says so. When you reject a proposal, the system learns — structurally — through the feedback loop.
The kill switch exists. The audit trail is non-negotiable. Mutations require your explicit hand. We built it this way because federal docket work demands it.
APEX is in closed beta with a small number of administrative law tribunals. If you run a docket and want to evaluate it for your office, write us. We'll respond personally.