01 / 08
QuNetra · Decision Infrastructure
Industry Architecture Analysis

The Legal AI Stack
Has a Governance Gap

AI can now participate in legal execution — drafting, reviewing, routing, and acting inside workflows.

But who governs the action itself?

02 / 08
The Baseline

The Traditional
Legal Stack

Decades of legal technology converged into systems built to record, organize, and reconstruct work.

Financial & eBilling Elite · Aderant · Thomson Reuters · Wolters Kluwer
Document Management iManage · NetDocuments
Intake, Conflicts & Risk Intapp
Litigation & eDiscovery Relativity · Reveal
Matter Management Various ELM / matter platforms
Built to record and manage legal work — not to govern autonomous execution.
03 / 08
The Shift

The New
AI Layer

A new tier of platforms now reasons, retrieves, and increasingly acts inside legal workflows.

AI Reasoning & Drafting Harvey · Hebbia · Microsoft Copilot
Retrieval & Knowledge Reasoning Hebbia · emerging legal-vertical platforms
Workflow & Orchestration ServiceNow · Intapp · custom platforms
Legal Agents & Operational AI Emerging cohort across drafting, intake, review
AI is moving from assistance toward operational participation.
04 / 08
The Execution Problem

Some Actions
Cannot Be Undone

When AI participates in legal execution, certain actions create immediate operational consequence.

Filing
Position taken
A submission to court or regulator that cannot be silently retracted.
Billing Entry
Revenue claim made
A financial assertion that becomes an audit and compliance artifact.
Document Edit
State changed
A material change to a working document or matter record.
Disclosure
Information released
A communication that cannot be un-shared once it has occurred.
Once committed, operational consequences already exist.
05 / 08
The Distinction

Audit Execution Governance

Two operational models. Different timing. Different control surface.

Auditability
After the action
  • Retrospective
  • Explanations & reconstructions
  • Defensible record
  • Cannot prevent commitment
Governed Execution
Before the action
  • +Runtime validation
  • +Admissibility controls
  • +Defensible execution
  • +Evidence in line with the act
Authorization before execution and audit after execution are fundamentally different operational models.
06 / 08
The Missing Layer

Decision Infrastructure

The runtime gate between decision and execution — where action is revalidated, evidenced, and committed under governance.

01 Runtime Governance Re-checks the action against current state, policy, and authority at the moment of commit.
02 Admissibility Validation Asks: is this action operationally admissible right now?
03 Evidence at Execution Generates immutable evidence in line with the act, not reconstructed afterward.
04 Governed Execution Issues a runtime verdict that determines whether the action proceeds.
ALLOW HOLD DENY ESCALATE
07 / 08
The Reference Architecture

The Future
Legal Stack

Five layers. Decision Infrastructure is the governance layer that determines whether the others operationalize safely.

L1 AI Reasoning Drafting · Retrieval · Copilots
L2 Workflow & Orchestration Routing · Approvals · Coordination
L3 Knowledge & Data Documents · Metadata · Provenance
L4 Matter & Financial Systems Billing · Matters · Records
L5 Decision Infrastructure Runtime Governance Layer
Four layers are maturing fast. The governance layer determines whether they can be safely operationalized.
08 / 08
The Architectural Shift

Governance that exists only after execution is no longer sufficient.

The future legal stack governs execution itself.

QuNetra.ai
Vision Beyond Classical
Explore the architecture shift
qunetra.ai/blog/legal-ai-governance-gap