04 — Legal Twin
Legal as a lever, not a gate.
A model of the contract estate — obligations, deviations and cycle time — that turns review into a measured, accelerating step in revenue.
North Star
Cycle time to signature · Value at risk
The stream
Automation → Capability → Value driver → Financial value
Automation
First-pass clause review and redlining
Capability
Standard contracts reviewed in minutes
Value driver
Deal velocity
Financial value
Revenue recognized earlier
Automation
Obligation extraction and tracking
Capability
Know every commitment made
Value driver
Risk exposure
Financial value
Avoided penalties and leakage
Automation
Playbook-driven negotiation support
Capability
Consistent positions across regions
Value driver
Standardization
Financial value
Lower external counsel spend
Value system
Value ≠ North Star ≠ KPI ≠ driver ≠ process metric.
The translation layer that connects a corporate objective to something an operating team can actually move on Monday.
Value
Contract velocity without exposure
Economic outcome
Revenue recognized earlier, claims avoided
North Star
Contract cycle time at constant risk score
Outcome KPIs
Driver metrics
Process metrics
Causal model
Value leakage
Where the value goes missing today.
Leakage is multiplicative. Every gate that stays leaky discounts everything built upstream of it.
Intake
Requests arrive without the facts needed to draft.
Cycle days
Review
Standard clauses re-negotiated deal by deal.
Counsel capacity
Escalation
Everything escalates because thresholds are informal.
Delay
Obligations
Signed commitments never tracked post-signature.
Unpriced risk
Digital twin
Simulate the interventions before anyone funds them.
Instead of implementing ten recommendations and learning the result in six months, the twin stacks them first.
Contract cycle time
Guardrails
- Risk score per contract not above baseline
- No deviation outside approved playbook
- Counsel sign-off on all non-standard terms
Constraints
- Jurisdictional templates
- Insurance and liability caps
- Retention and privilege rules
Automation
What runs without asking.
Intake triage
Routes by risk tier; auto-approves within playbook.
Clause Review Agent
Redline generation
Proposes fallback positions with rationale.
Redline Negotiation Agent
Obligation register
Live register of commitments, dates and owners.
Obligation Tracking Agent
Clause Review Agent
Flags deviations and quantifies value at risk.
SLM extraction + LLM reasoning
Redline Negotiation Agent
Drafts positions from precedent, never invents terms.
LLM + RAG on precedent library
Obligation Tracking Agent
Monitors and escalates commitments.
SLM classification + rules
Integrations
Where it plugs into the domain.
Vendor-agnostic by design. The twin reads and writes through whatever stack the domain already runs on.
Systems of record
- CLM
- Document management
- Matter management
Workflow
- E-signature
- CRM / CPQ
- Ticketing and approvals
Data & runtime
- Precedent library (RAG)
- MCP connectors
- Policy engine
Domain platforms we connect
Contract lifecycle
Signature & identity
Matter & spend
Knowledge & research
Entity & IP
Data & runtime
Plus anything else with an API — connectors are added per engagement, not sold as a platform lock-in.
Value
The levers, and what they move.
Speed
Remove queue time from the revenue path.
Days to signature
Cost
Insource routine review.
External counsel spend
Trust
Consistent, defensible positions.
Deviation rate · Value at risk
Risk
What could go wrong, and what stops it.
Every risk in this domain has a named containment in the runtime — not a slide.
Hallucinated clause or citation
Retrieval-grounded drafting; every clause traced to the playbook
High
Privilege or confidentiality breach
Matter-level access control; privileged corpus isolated
High
Unauthorized commitment by an agent
Agents draft only — no signature or binding authority
High
Compliance · Security
Built into the runtime, not bolted on.
EU AI Act
Limited risk — professional-use transparency with mandatory human legal sign-off.
- No agent output is legal advice without qualified human approval
- Citation of source precedent for every proposed clause
- Model and version recorded on each reviewed document
GDPR
- Confidentiality and privilege boundaries enforced by matter
- Counterparty personal data minimized before embedding
- Retention aligned to matter closure policy
Security
- Matter-level access control
- Private, non-training inference
- Watermarked exports and audit of downloads
Controls & oversight
- Playbook thresholds by risk tier
- Escalation to counsel on any novel clause
- Immutable review history
Agentic execution
Agents earn authority. They are not given it.
No agent in this twin controls anything it has not first proven in replay, evaluation, simulation and shadow.
01
Historical replay
Re-run the last 12 months. Would the agent have been right?
02
Offline evaluation
Scored against held-out outcomes, not opinion.
03
Digital twin
Simulated against the causal model under stress.
04
Shadow mode
Runs live, decides nothing. Divergence is logged.
05
Human recommendation
Proposes; a person executes and rates it.
06
Bounded pilot
One segment, capped exposure, hard rollback.
07
Human-supervised execution
Acts inside thresholds, humans approve exceptions.
08
Progressive autonomy
Authority widens only where evidence widened.
The unit of value
Don't buy transformation. Buy measurable movement.
This twin is contracted the way it is engineered: a baseline, a target, a guardrail, and an attribution method agreed in advance.
Baseline
Cycle time 24 days
Target
10 days at equal or lower risk score
Guardrail
Deviation rate and claim frequency held flat
Proof / attribution
CLM timestamps plus independent risk scoring of signed contracts
Next twin
Compliance Twin →