Make enforcement expertise available to the whole team
Lex is an AI agent: software that can plan and carry out a sequence of tasks, such as searching sources and drafting summaries, within rules the team sets. Today the know-how for enforcement sits with a few experienced people. Lex is meant to make it shared and repeatable.
Run approved searches for protected marks and record what was and was not covered.
Match new findings to past cases, IPMS records, and court outcomes.
Draft briefs, correspondence, and next steps for a person to review and approve.
What changes in a day's work
Much of the time spent on a matter goes to gathering and reconciling information rather than judging it. Lex is meant to reduce that work so people can spend more time on judgment.
Today
- Search websites and public sources.
- Check several internal systems and historical files.
- Cross-check IPMS (our intellectual property management system), email, and past case records by hand.
- Confirm that parties, marks, and case status match.
- Prepare evidence, summaries, correspondence, and follow-ups.
- Reconstruct the reasoning when management asks.
With Lex (target)
- Set the target and approved scope.
- Review sourced findings and a list of what was not checked.
- Review one case brief with open questions flagged.
- Approve next steps; Lex prepares the drafts.
Today is six manual steps; the target is four. The right-hand column is the target: the case brief and drafting steps are not yet built.
Three sources of knowledge
Investigation methods, evidence practices, and legal standards. Legal points must be checked against sources for each jurisdiction.
Team procedures, case histories, court outcomes, and approved templates.
Public search results today. Later, current IPMS case status, owners, and deadlines.
Standard software enforces permissions, spending limits, record accuracy, and audit trails. The AI interprets unclear information, judges relevance, and proposes next steps for people to review.
The team teaches Lex how Saba works
The proposed process turns reviewed experience into shared knowledge. It does not turn every opinion into a rule.
Case, procedure, correction
Expert validation
Named owner and version
Checked against test cases
| Team contribution | How Lex uses it | Governance |
|---|---|---|
| Investigation procedure | Follows a repeatable, approved method | Named owner, version, effective date |
| Previously handled matter | Retrieves context and compares a new lead | Source record, access controls, up-to-date check |
| Correction to a finding | Improves rules, instructions, or test examples | Reviewer's rationale and approval |
| Judgment or settlement outcome | Shows relevant historical context | Confirm identity and legal scope; never assume it applies automatically |
| Recurring exception | Suggests a documented escalation or process change | Confirmed across multiple cases before adoption |
From possible infringement to case
Run bounded text searches against protected marks, keep a history of each run, and report gaps.
Is this lead new, already handled, previously reviewed, or a repeat? Show the matching evidence and what remains uncertain.
Bring together evidence, related matters, missing facts, options, and draft correspondence for approval.
Track assigned tasks, approvals, deadlines, and outcomes through authorized systems.
Illustrative case: BlackRock
Suppose Lex finds a new BlackRock-related listing. A future case-history feature could surface a related past Saba matter and its recorded court outcome. Lex would compare parties, marks, conduct, jurisdiction, dates, and the scope of the judgment, then flag open questions for review.
It must not conclude that the new listing is already resolved just because Saba won a related case. The actual records and the scope of that judgment must be verified first.
Responsibility with limits, not unchecked authority
| Activity | Rule |
|---|---|
| Search approved public sources | Automatic, within approved scope and limits |
| Interpret and prioritize leads | AI-assisted, backed by evidence, open to review |
| Link findings to past cases | Proposed link with a stated confidence level; a person confirms unclear matches |
| Prepare briefs and correspondence | Draft only, using approved records and templates |
| Send notices, start proceedings, or change official records | Separate permission and explicit human approval required |
| Change team procedures | Governed review and release process |
When Lex misses something
Lex should show what it searched, what it skipped, and what it could not reach. A missed case should trigger a documented review of the search terms, sources, and follow-up steps. Finishing the planned search does not prove that nothing was missed.
Lex's responsibilities grow in stages
A stage opens only after its tests are passed. Each stage adds responsibility and requires more evidence.
| Stage | Responsibility | Proof required to advance |
|---|---|---|
| 1 · Supervised researcher | Run bounded searches and report sources and gaps | Reliable, complete execution and accurate reporting of coverage |
| 2 · Evidence-aware investigator | Identify relevant leads and prepare evidence for review | Reliably finds known relevant cases; every finding has its sources recorded and tested |
| 3 · Case-aware analyst | Connect findings to Saba's case history | Accurate matching to past cases; up-to-date records |
| 4 · Operations coordinator | Prepare approved work items, drafts, and follow-ups | Role-based permissions, complete audit trails, approval workflows |
| 5 · Supervised operational teammate | Coordinate multi-step work and handle exceptions | End-to-end testing, a named operational owner, ongoing governance |
Current position: early Stage 1. The next milestone is showing that Lex reliably finds known relevant cases, which has not yet been demonstrated.
Measure outcomes, not just activity
Share of known relevant cases Lex finds, share of flagged leads that prove useful, how often reviewers agree, and whether each finding has its source.
Completes approved work, reports gaps honestly, stops when asked, and keeps a complete audit record.
Review time saved, duplicate research avoided, accuracy of case matching, and timeliness of follow-ups.
Testing will include known relevant and irrelevant examples, difficult cases, incomplete data, and disputed findings. Targets are set by Legal & Enforcement leadership.
A foundation other agents can reuse
Lex is the first agent built on ADI's shared controls: knowledge governance, tool access, approvals, testing, and audit trails. The same controls can support agents for other teams, with their own knowledge and rules.