Change Management Architecture

1. Module Name and Identifier

Name: Change Management Architecture for Legal AI Adoption

Code: TAL-03

Pillar: Pillar 3 — Talent, Literacy & Change

Primary Layers: Execution Layer (E), with Measurement (M) and Optimisation (O) interfaces

Methodology Version: v2026.1

Verification Date: 2026-05-23

2. Purpose and Scope

This Module defines the institutional change architecture that converts AI availability into AI use. It provides the disciplines, artefacts, and cadence required to:

• Move legal functions from ad-hoc tool deployment (Band 1) to structured adoption (Band 2+).
• Evidence adoption under the Defensible AI standard, specifically for the DPS Adoption lens.
• Ensure AI capabilities are integrated into daily legal workflows, not left as unused inventory.

In scope:

• Human-side change disciplines for AI adoption (sponsorship, stakeholder engagement, champions, workflow redesign, training, recognition).
• Execution-layer operating model for AI-enabled workflows.
• Evidence production for DE-2 (Methodology transparency) and DE-5 (Continuous learning).
• Adoption telemetry and ROAI (Return on AI) measurement.

Out of scope:

• Vendor selection and procurement.
• Core governance policy drafting (handled by GOV-series Modules).
• Technical model evaluation and benchmarking (handled by GOV-09 / evaluation Modules).
• Band 4–5 sustaining patterns (handled by SUS-05).

3. Problem Statement and Rationale

Legal functions increasingly acquire AI tools but fail to achieve meaningful adoption:

• Licences are purchased and pilots run, but practitioner behaviour and billable hours do not shift.
• The AI Bill of Materials (BoM) grows while the Adoption lens of the Maturity Stack remains flat.
• Boards, regulators, and Editorial Councils expect evidence of operating use, not just procurement.

Without a formal change architecture:

• AI remains a tool inventory, not an operating posture.
• The function cannot produce DPS Adoption-lens evidence.
• Shadow AI (Class 6) proliferates as practitioners self-solve outside governance.

This Module closes the gap by sequencing the human-side disciplines that turn AI availability into AI use and by standardising the artefacts that demonstrate defensible adoption.

4. Position in Legal AI OS (Pillars, Layers, Lenses)

Pillar Alignment:

• Primary Pillar: Pillar 3 — Talent, Literacy & Change.
• Role: Institutional capability to absorb new operating models (AI as a primary case).
• This Module is the formal change capability within Pillar 3 that enables scalable absorption.

Layer Alignment:

• Execution Layer (E): Primary operating layer — where practitioners use AI in daily work.
• Strategy Layer (S): Consumes the AI vision and sponsorship intent defined elsewhere; this Module operationalises that intent.
• Governance Layer (G): Provides constraints (GOV-02, GOV-03, GOV-04, GOV-08, GOV-14, GOV-16) that inform workflow design and training.
• Measurement Layer (M): Receives adoption telemetry and ROAI metrics.
• Optimisation Layer (O): Uses operating-blocker intelligence from champions and dashboards.

Maturity Stack Lenses:

• Primary lens: Adoption.
• Secondary lens: Sophistication.

Band Transitions:

• Band 1 → Band 2 (Foundational → Operational):
  • From: Tools available; usage ad-hoc; no systematic telemetry.
  • To: Tools in use by ≥40% of named practitioners; structured adoption practice in place.
• Band 2 → Band 3:
  • From: Tools in use.
  • To: Tools embedded in primary legal workflows with explicit review gates and risk-class assessments.
• Band 3 sustaining: Adoption durable across leadership changes; Module run as a quarterly cycle.

5. Defensibility and Risk Taxonomy Mapping

Defensibility Elements (DE):

• DE-2 — Methodology transparency
  • Evidence: AI vision statement; communication architecture; stakeholder maps; workflow maps with AI integration points; risk-class assessments; champion roster and responsibilities.
• DE-5 — Continuous learning
  • Evidence: Champion network operations; training cadence and records; adoption and ROAI dashboards; Shadow AI incident logs; quarterly and annual refresh cycles.

Risk Taxonomy 2026 Coverage:

The Module touches nine classes, with emphasis on:

• Class 1 — Hallucination: Review gates and sampling protocols in workflow design.
• Class 2 — Data leakage: Integration with DAT-03 and safe-handling practices in training.
• Class 3 — Bias: Bias-testing review gates per GOV-04 in workflow design.
• Class 4 — Vendor lock-in: Backup processes and portability in workflow design.
• Class 5 — Regulatory non-compliance: Jurisdictional checks at workflow design; refresh on regulatory change.
• Class 6 — Shadow AI: Champion network as primary detection and reporting mechanism; Class 6 register entries feeding GOV-03.
• Class 7 — Client confidentiality: Privilege-preserving workflow design; client disclosure language.
• Class 8 — Professional conduct: ABA 5.3 supervising-lawyer responsibilities embedded in workflows and training.
• Class 9 — Accountability dilution: Explicit role-based accountability for AI-enabled outputs.

Agentic Tier (Level 4) Specifics:

• Mandatory Agentic Tier briefing for all champions before any Level 4 deployment is communicated.
• Delegation-Authority Register entries per GOV-14.
• Materiality Calibration per GOV-16.
• Evidence Register and oversight cadence per GOV-13 and GOV-15.
• Explicit logging of deployment resistance and concern themes.