Matter tracker
Internally built in Smartsheet or Excel; out of date within a week of board day.
CaseOps for general counsels
The operating layer for in-house legal.
Portfolio, counsel, contracts, obligations and compliance — one matter graph that gives the GC a real answer when the CEO asks, 'what are we exposed to?'
1
Portfolio view
Every matter, counsel and spend line on one surface.
Audit export
Board preparation
Date-scoped records and retained activity history.
Tenant-scoped
Access controls
Authorization is checked against the firm's records and permissions.
What's broken today
Five tools, none of them agreeing.
A GC stitches a matter tracker, a contract repo, an outside-counsel spreadsheet, a compliance calendar and a board deck. Nothing rolls up. The quarterly board answer is improvised.
Contract repo
A DMS folder. Obligations extracted manually by a paralegal if at all.
Outside counsel spend
Email invoices, a quarterly consolidation, no realisation against budget.
Compliance calendar
A PDF owner's list; the owner left two quarters ago.
The real cost: every board meeting becomes a fire drill. The GC can answer the question, but only by paying a human week to reconcile four systems. CaseOps replaces the reconciliation, not the humans.
The AI angle
Explainable answers the board will accept.
AI compresses 80 matters and 200 contracts into 'here is what is open, here is who said what, here is what is due next quarter' - with every number traceable to a source. Litigation intelligence stays source-backed, reviewable, and explicitly non-advisory.
Contract → obligation — shipped
LLM extracts parties, covenants, payment terms, consent clauses, audit rights today. Every duty becomes a task; due dates pulled where the contract states them, placeholders where it doesn't.
Portfolio rollups
AI summarises 80 matter records into a board-ready extract. Every line is traceable to the underlying matter and audit event.
Grounded recommendations
Forum, authority and next-action recommendations with rationale, assumptions, missing facts, and confidence on every option.
Litigation Intelligence review
Proceeding directions, affidavit gaps, mock-hearing feedback, calibrated historical patterns, and knowledge-graph relationships appear with source links and review state.
AI policy + ModelRun audit — shipped
Tenant-scoped tenant_ai_policy gate enforced at runtime on the structured-call path (matter summary, draft preview, recommendations, hearing pack). Every LLM call writes a ModelRun audit row with provider/model/tokens/latency/status — visible per matter.
What the GC gets to say
"Of our 83 open matters, 11 have a deadline in the next 30 days. Of our 214 active contracts, 6 have audit rights expiring this quarter and 3 have payment obligations overdue. Here is the audit export covering every AI action that went into those numbers."
Agentic posture: AI assistance is review-first: cause-list candidates, obligation-due suggestions, and intake triage stay behind human approval. Autonomous scoped-agent execution is planned, not live. Cross-tenant training is off by default.
Portfolio
Every matter on one surface, rolling up to the board.
A matter is the unit of work. Its stage, documents, counsel, spend and obligations all attach to the same record. The portfolio view is the same record, aggregated.
Single matter graph
Open, pending, on hold, closed — with stage, forum and next action in one row per matter.
Live stage + next action
No more 'last update on Tuesday'. The cockpit reflects state the moment anyone records an event.
Team scope and walls
Restrict visibility at a matter level when a case is conflicted or sensitive — without reshuffling roles.
Outside counsel
Brief, budget and measure — without a spreadsheet.
Outside counsel becomes a first-class workspace. Assign a matter to a firm, set a budget with alerts, see realisation against it, compare outcomes across firms.
Directory with outcomes
Rate cards, practice areas, historical outcomes, conflict flags — not just a contact list.
Budget with alerts
Spend cap per matter. Invoices beyond the cap require a GC override, recorded on the audit trail.
Realisation by firm
See billed, collected and aging per counsel. Compare realisation rates across firms before the next brief.
Outside-counsel portal — shipped
Every briefed firm gets a magic-link login at /portal/oc — see assigned matters, upload work product, submit invoices, log time. Scope-gated by MatterPortalGrant; the GC sees every action on the audit trail.
Client portal + KYC — shipped
The client gets their own /portal view of the matter (Overview, Comms, Hearings, KYC tabs). KYC submit + GC review + audit on every state transition. Built on the same Phase C portal scaffold; no second auth system to operate.
Statutes attached to the matter
Attach available source-verified provisions to the matter and inspect their exact source version. Catalog coverage remains incomplete; unverified entries are not selectable as verified references.
Board reporting
Review assigned matters, outside-counsel fees and budget use from the recorded portfolio. Keep the underlying invoices and activity history available for reconciliation.
Evidence limits
Reports reflect the data entered and reconciled in the workspace. Missing invoices or activity records are not evidence of lower legal spend.
Contracts and obligations
Pull duties out of contracts. Track them where they live.
Store contracts, review AI-extracted clauses, track obligations and compare against a playbook. Extraction depends on document quality and provider availability; review the source contract before relying on the result.
Clause extraction
Review extracted parties, term, payment, confidentiality, IP, liability and other available clauses alongside the source contract. Missing or uncertain clauses require review.
Obligation tracker
Payment, reporting, notice, renewal and termination duties become tasks with due dates where the contract states them. Surfaces on the matter cockpit.
Playbook compare
A 15-rule default Indian commercial playbook seeds in one click. The OpenAI review model compares each clause against the expected position and flags matched / missing / deviation with severity.
Risk, audit and AI posture
Control is built in, not bolted on.
The standard a GC should expect: tenant isolation by construction, matter-level ethical walls, audit on sensitive actions, AI policy enforcement, malware-scan guardrails, and documented rotation runbooks. Enterprise SSO/SCIM, autonomous agents, and dedicated provider connectors remain planned or provider-gated until readiness evidence is complete.
Tenant isolation at the engine layer
Every record carries a tenant id. Storage, query and retrieval filter by it, not only the application layer.
Matter-level ethical walls
A walled matter is invisible to users outside the wall — including admins. Override, not weaken.
Audit on every sensitive action
Upload, draft, approval, AI run, recommendation accept, invoice state change — recorded with actor and before/after state.
AI policy enforced at runtime
Tenant-scoped tenant_ai_policy gate runs on every structured AI call (matter summary, draft preview, recommendations, hearing pack). Blocked-model attempts return HTTP 403 with the specific model that was rejected, and a ModelRun audit row records the refusal.
When the board asks for evidence
Export the audit trail for a date range as JSONL or CSV. Every sensitive action is recorded with actor, timestamp and before/after state. Recommendation acceptances are distinguishable from human actions.
Cross-tenant training is off by default. Your matters and contracts are not pooled for model training without an explicit written opt-in.
Research evidence
Sources you can inspect.
Review the cited source and its available coverage. Corpus size and retrieval quality are different measures; a self-recall probe is not a guarantee of legal relevance.
Source-linked
Judgment references
Open the cited authority and inspect its provenance.
Voyage
Embedding pipeline
Production retrieval uses voyage-4-large embeddings.
Reranked
Research results
Cross-encoder ranking supports source-based review.
Not certified
Corpus quality score
No representative legal-retrieval rating is claimed here.
Tenant-private by default
Your matter documents never leave your workspace for training. Separate from the public authority corpus.
Source-aware legal review
Inspect citations, source coverage and limitation notes before relying on generated work. Legal review remains necessary.
Contact
A 45-minute walkthrough, shaped to your sector.
We set up a sandbox with a sample portfolio from your industry — banking, SaaS, pharma, infrastructure — and walk through outside-counsel spend, obligations and the audit export with your team.
Direct contact
sanjeev@orchestrum.inWrite directly for a walkthrough, a sector-specific sandbox or a security and DPA conversation. Founder-led until we are larger.