Legora Agentic Law Transformation Framework

// When should a legal team use the Legora Agentic Law Transformation Framework?

Use this skill when a legal team, law firm, or in-house legal department is evaluating, deploying, or scaling AI — particularly when moving beyond first-generation 'Legal AI' efficiency tools toward autonomous, agentic execution of legal work. Trigger when stakeholders ask 'what's next after we've automated the basics?' or when the current AI deployment feels like faster-same-work rather than fundamentally different work.

// What information do you need before applying the Legora Agentic Law Framework?

• Legal team typerequired
  Is this an in-house legal team, a law firm, or a legal services business? What practice areas or departments are involved?
• Current AI maturityrequired
  What AI tools are already deployed? What workflows have been automated? What is the team's experience level with AI adoption?
• Team size and structurerequired
  How many lawyers, what headcount, how many jurisdictions, how many clients or internal stakeholders served?
• Primary pain pointsrequired
  What work is high-volume and low-complexity? Where are bottlenecks? What client or stakeholder service gaps exist?
• Strategic ambitionrequired
  Is the goal efficiency, competitive differentiation, or full reinvention as an AI-native legal team?
• Existing tech stack
  What document management systems, data sources, signing platforms, or third-party publishers are already in use?
• Regulatory or jurisdictional scope
  Which jurisdictions and regulatory domains does the team operate in? Relevant for research and compliance monitoring design.

// What are the core principles behind the Legora Agentic Law Transformation Framework?

// How do you apply the Legora Agentic Law Transformation Framework step by step?

1. 1

   Diagnose which of the Three Phases of Agentic Law transformation the legal team currently occupies

   Phase 1 — Leverage: AI handles high-volume, low-complexity work. Lawyers get capacity back. Matters move faster. Most organisations are here. Phase 2 — Differentiation: Service models adjust to the type of work required. A team of five competes like a team of 50. New value functions and client relationships emerge. Competing on service delivery innovation, not headcount. Phase 3 — Reinvention: AI-native legal teams rethink processes from the ground up. Knowledge is productised and deployed at scale — automatic intake, triage, responses to legal queries. New business models emerge. This is still new territory.

2. 2

   Map the current tech stack against the six layers of the Agentic Operating System (AOS)

   Layer 1 — Foundation Models: What LLMs power current tools? Are they horizontal (general purpose) or vertical (legal-specific)? Layer 2 — Agentic Harness: Is there a harness giving the model memory, guardrails, multi-step execution, and tool routing? Or is the team using stateless prompting? Layer 3 — Data and Integrations: What document management systems, signing platforms, data sources, and third-party publishers exist? Are they unified or siloed? Layer 4 — Context and Knowledge: Are precedents, clause libraries, playbooks, and matter history stored and accessible to the AI? Layer 5 — Legal Skills and Capabilities: Are domain-specific workflows, instructions, and long-context handling capabilities built in? Layer 6 — Enterprise Security and Governance: Are ethical walls, matter-centricity, audit trails, and permissions in place? Identify which layers are missing, weak, or disconnected.

3. 3

   Identify the highest-value use cases for agentic execution in this specific legal team

   Prioritise by: volume (high-frequency tasks), complexity (low-to-medium complexity first), pain (where time is most wasted — e.g. data room reorganisation, closing checklists, due diligence report drafting, regulatory monitoring). Use the following archetypes to prompt thinking: (a) Unstructured data room organisation — sorting, renaming, custodian attribution at scale; (b) End-to-end due diligence — multi-workstream, multi-agent parallel execution across corporate, IP, employment, real estate, data protection, tax; (c) Transfer pricing / intercompany structure visualisation — corporate structure diagrams, risk heat maps, radar charts; (d) Regulatory horizon scanning — jurisdiction and topic-specific monitoring with gap analysis against existing policies; (e) Closing checklist generation and management — populated from credit agreements or transaction documents, with live status tracking.

4. 4

   Design the Skills architecture — encode the team's knowledge and processes into plain-English agent instructions

   Skills are custom instructions that extend what the agent can do. Write Skills as if briefing a new team member in plain English. The agent proactively loads the right Skill based on the task. Skills should capture: how the team structures its work product, firm-specific language and style guides, practice-area-specific workflows, risk tolerance rules and escalation triggers, and client-specific preferences. Lawyers are the best prompt engineers in the world — involve them in Skills authorship. Offer a library of out-of-the-box Skills for common legal work, then customise.

5. 5

   Implement Lists as the connective tissue between agents, lawyers, and stakeholders

   Lists are how work gets organised, assigned, and executed inside the agentic OS. Every matter has structure: tasks, dependencies, handoffs, deadlines. Lists make this structure visible and executable. Use cases: closing checklists, due diligence trackers, case chronologies, in-house intake. Each list item is an atomic unit — an object with metadata (status, assignee, parties, jurisdiction, source citation, start/end dates). The agent can populate, update, and flag items autonomously. Humans review, annotate, and reassign. When the agent completes a task inside a list, the work product returns to the list for review and the relevant lawyer is notified. Lists replace Excel, Word, and physical Post-its — none of which carry structure.

6. 6

   Design the human-agent collaboration model with oversight checkpoints

   For any long-running agentic task: (a) Set the objective clearly; (b) Agent produces a Plan before executing — review and approve or modify it; (c) Agent flags uncertainty and asks clarifying questions mid-task rather than guessing; (d) Agent returns completed work product to the list or surface for human review; (e) Legal Engineers support deployment, tailoring, and ongoing optimisation. Never design a workflow where the agent runs to completion without human checkpoints — the plan approval moment is critical for setting intent.

7. 7

   Set up Monitors for proactive regulatory horizon scanning relevant to the team's jurisdictions and practice areas

   Select jurisdictions and topics (e.g. EU antitrust, UK financial services, employment law). The monitor tracks relevant regulatory changes in real time across official sources, updated hourly. The agent stitches regulatory changes together with the firm's existing policies and documents to run gap assessments, compliance briefs, or policy updates — all cited to original sources. For law firms: enables a managed service and deeper client relationships previously too cost-prohibitive. For in-house teams: real-time compliance monitoring at scale, previously impossible. Output should include: key regulatory developments, what changed, why it matters, what clients should do, key dates and deadlines — formatted for board briefings or client updates.

8. 8

   Define the new economic and service delivery model that Agentic Law makes possible

   Ask: What service could a team of five deliver that previously required a team of 50? What static guides, handbooks, or market standards could become AI-powered workflows accessible 24/7? What legal knowledge could be productised and deployed at scale (automatic intake, triage, query response)? What managed services or subscription-model relationships become viable? Competitive advantage in the Agentic Law era comes from service delivery innovation, not headcount. Reframe the team's value proposition accordingly.

9. 9

   Build the change management plan and identify or develop Legal Engineer capability

   AI transformation is not linear and does not happen overnight. Change management is as important as technology. Identify who inside or alongside the team understands both AI capabilities and legal workflows, risk tolerance, and organisational culture. If no Legal Engineers exist internally, plan for their development or external partnership. Deploy change management support in proportion to technology deployment — not as an afterthought.

// What are real-world examples of the Legora Agentic Law Framework in action?

// What mistakes should you avoid when implementing Agentic Law transformation?

• Applying the same model, structure, and ways of working that already exist to new AI tools — building something smarter and faster but not fundamentally different. This is the Legal AI trap.
• Using horizontal, general-purpose LLMs with legal plugins and treating them as equivalent to a deep vertical legal system. A platform built for everyone is built for no one.
• Sending an agent off to complete a long-running task without a plan approval checkpoint at the start — failing to set the right intent with the model from the outset produces poor outcomes.
• Expecting AI transformation to happen overnight or to follow a linear path — it typically happens in three large phases and requires sustained change management alongside technology.
• Delivering work product in static Excel files or plain Word documents — failing to package outputs in a way that communicates value undermines stakeholder perception of what the team has achieved.
• Treating technology deployment as sufficient without investing equally in Legal Engineer capability — change management is as important as the technology itself.
• Designing a workflow where the agent runs entirely to completion without human oversight checkpoints — the agent must surface uncertainty and involve human judgment, not simulate judgment it is not fit to make.
• Confusing a raw foundation model (stateless, no memory, no guardrails) with an agentic harness — without the harness, the model has no spine and cannot carry a task through multi-step execution.
• Building point solutions for narrow problems rather than a single connected system — individual tools cannot achieve the compound effect that comes from all layers working together under a single roof.

// What do key terms like AOS, Skills, Lists, and Monitors mean in the Legora framework?

Agentic Law

The next era of legal technology — a fundamentally different model for how legal work gets done, where AI agents autonomously plan, execute, review, and complete legal work end-to-end with human oversight integrated throughout. Distinct from Legal AI, which only made lawyers faster at the same work.

Legal AI

First-generation AI applied to legal work — smarter and faster execution of existing workflows, but not fundamentally different in structure or model. Pronounced dead as a sufficient destination; it is now a foundation, not an endpoint.

Agentic Operating System (AOS)

A single connected system that facilitates the flow of information, communication, and execution of legal work across six engineered layers: Foundation Models, Agentic Harness, Data and Integrations, Context and Knowledge, Legal Skills and Capabilities, and Enterprise Security and Governance.

Agentic Harness

The layer that gives a raw LLM its spine — memory, guardrails, tool routing, and the ability to carry a task through multi-step execution (retrieval, checking, flagging, drafting, reviewing) all the way through to completion. A raw LLM without a harness is stateless and insufficient for legal work.

Legal Engineer

A role unique to the Agentic Law era — former lawyers who are deep AI experts, understanding the workflows, risk tolerance, and culture of change inside legal organisations. Their sole purpose is to accelerate the realisation of value by deploying, tailoring, and partnering with legal teams to bring the AOS to life.

Skills

Custom instructions written in plain English that encode a legal team's knowledge, processes, and standards into the AOS, extending what the agent can do. The agent proactively loads the right Skill based on the task. Lawyers are described as the best prompt engineers in the world for authoring Skills.

Lists

The connective tissue between the agent, the lawyer, and other stakeholders inside the AOS. Lists organise, assign, and execute work — every matter's tasks, dependencies, handoffs, and deadlines made visible and executable. Each list item is an atomic unit (an object) with structured metadata, not just text.

Monitors

A proactive regulatory horizon scanning capability — the user selects jurisdictions and topics, and the agent tracks relevant regulatory changes in real time across official sources, stitching changes together with the team's existing policies to produce gap assessments, compliance briefs, or policy updates, all cited to original sources.

Three Phases of Agentic Law Transformation

Phase 1 — Leverage: AI handles high-volume, low-complexity work, giving lawyers capacity back. Phase 2 — Differentiation: Service models adjust, enabling a team of five to compete like a team of 50, winning on service delivery innovation not headcount. Phase 3 — Reinvention: AI-native legal teams rethink processes from the ground up, productising knowledge, deploying at scale, and creating new business models.

Knowledge Productisation

The Phase 3 activity of encoding a legal team's accumulated knowledge, processes, and standards into the AOS as Skills and automated workflows — making expertise scalable and deployable on demand, rather than locked in individual lawyers' heads or static documents.

Sub-agents

Specialised agents launched by a main orchestrating agent to work in parallel on distinct workstreams of a complex task (e.g. corporate, IP, employment, real estate workstreams of a due diligence). Sub-agents complete their portion and return their work product to the main agent, enabling work that would take six hours sequentially to complete in parallel.

Plan Approval

A critical human oversight checkpoint before an agent begins a long-running task — the agent presents its proposed plan of action, and the human reviews, modifies if needed, and explicitly approves before execution begins. Sets the right intent with the model from the outset.

// FREQUENTLY ASKED QUESTIONS