Frequently Asked Questions About Legora Agentic Law Transformation Framework
22 answers covering everything from basics to advanced usage.
// Basics
What does 'Legal AI is dead' actually mean?
It means first-generation Legal AI — tools that made lawyers faster at the same work — is no longer a sufficient destination. It made legal work smarter and quicker but not fundamentally different. The framework treats this as a foundation, not an endpoint. The real shift is to Agentic Law, a generational leap that changes what a lawyer can be, not just how fast they work.
What is the Agentic Harness and why does it matter?
The Agentic Harness is the AOS layer that gives a raw LLM its spine — memory, guardrails, tool routing, and the ability to carry a task through multi-step execution like retrieval, checking, flagging, drafting, and reviewing. Without a harness, an LLM is stateless and cannot complete legal work. This is the difference between a capable engine and an actual working car.
What are Monitors and how do they help legal teams?
Monitors are a proactive regulatory horizon-scanning capability. You select jurisdictions and topics, and the agent tracks relevant regulatory changes in real time across official sources, updated hourly. It stitches changes together with your existing policies to run gap assessments, compliance briefs, or policy updates — all cited to original sources. For firms this enables managed services; for in-house teams it enables real-time compliance at scale.
What are sub-agents in the Agentic Operating System?
Sub-agents are specialised agents launched by a main orchestrating agent to work in parallel on distinct workstreams of a complex task — for example, the corporate, IP, employment, and real estate workstreams of a due diligence. Each completes its portion and returns work product to the main agent. This lets work that would take six hours sequentially complete in parallel, dramatically compressing timelines.
// How To
How do I diagnose which AOS layers are missing or weak in my team?
Map your current stack against all six layers. Ask: Are your foundation models horizontal or legal-specific? Is there a harness giving memory and multi-step execution, or just stateless prompting? Are data and integrations unified or siloed? Are precedents and playbooks accessible to the AI? Are domain workflows built in? Are ethical walls, audit trails, and permissions in place? Flag any layer that is missing, weak, or disconnected.
How do I write effective Skills for my legal team?
Write Skills as if briefing a new team member in plain English. Capture how the team structures work product, firm-specific language and style guides, practice-area workflows, risk tolerance rules, escalation triggers, and client preferences. Involve lawyers directly — they're the best prompt engineers because they know the work. Start with a library of out-of-the-box Skills for common tasks, then customise to your firm's standards.
How do I set up a plan approval checkpoint for a long-running task?
Set the objective clearly, then require the agent to produce a plan before executing. Review, modify, and explicitly approve it before execution begins. During the task the agent should flag uncertainty and ask clarifying questions rather than guessing, then return completed work to the relevant List or surface for human review. Never let an agent run to completion without checkpoints — the plan approval moment sets intent.
How do I prioritise which use cases to automate first?
Prioritise by volume, complexity, and pain. Target high-frequency tasks that are low-to-medium complexity first, where time is most wasted. Strong starting archetypes include unstructured data room organisation, end-to-end due diligence with multi-agent parallel execution, corporate structure visualisation, regulatory horizon scanning, and closing checklist generation. These deliver fast, visible Phase 1 wins while building confidence for later phases.
How do I package legal work to communicate value?
Avoid delivering static Excel files or plain Word documents — they don't demonstrate value. Package outputs as interactive deliverables, structured portals, and visual outputs like structure diagrams, risk heat maps, and board-ready briefings. The framework's 'Work Packaging Matters' principle holds that perception of quality justifies the relationship; how you present the work is part of the value, not an afterthought.
// Troubleshooting
Why does my current AI deployment feel like faster-same-work?
Because you've likely applied new AI tools to existing models, structures, and ways of working — the 'Legal AI trap.' You built something smarter and faster but not fundamentally different. Escaping it requires rethinking the operating model itself: moving from augmenting tasks to agentic end-to-end execution, productising knowledge, and redesigning service delivery, not just accelerating what you already do.
My agent produces poor outcomes on complex tasks. What's wrong?
Most likely you sent the agent off without a plan approval checkpoint, so you failed to set the right intent from the outset. Require the agent to produce and present a plan before executing, and ensure it surfaces uncertainty and asks clarifying questions mid-task. Also check whether the Agentic Harness layer exists — without memory and multi-step execution, no model can carry a complex task through.
We deployed the technology but adoption stalled. What went wrong?
You likely treated technology deployment as sufficient without investing equally in change management and Legal Engineer capability. AI transformation is not linear or overnight — it unfolds across three phases and requires sustained change management proportional to the technology. Identify or develop people who understand both AI capabilities and your legal workflows, risk tolerance, and culture. Deploy as many Legal Engineers as software engineers.
Why are my individual AI tools not delivering compound value?
Because you're building point solutions for narrow problems rather than a single connected system. Individual tools can't achieve the compound effect that comes from all six AOS layers working together under one roof. Information doesn't flow, context isn't shared, and governance is inconsistent. The framework's remedy is a unified Agentic Operating System where data, Skills, Lists, and Monitors reinforce each other.
// Comparisons
How does Agentic Law compare to simply hiring more lawyers?
Agentic Law shifts competitive advantage from headcount to service delivery innovation. In Phase 2, a team of five can compete like a team of 50 by automating high-volume work and productising knowledge. Hiring scales cost linearly; agentic execution scales capacity without proportional headcount. The framework asks what a team of five could deliver that previously required 50 — that's the reframing that changes the economics.
How does a vertical legal AI system compare to a horizontal platform with legal plugins?
A horizontal platform built for everyone is built for no one. Legal work requires a deep vertical stack: domain-specific data, task management, legal interfaces, governance, security, workflows, integrations, and human partnership. A general-purpose AI with legal plugins lacks the harness, context, and governance layers legal work demands. The framework's 'Engine is Not a Car' principle captures exactly this distinction.
How does this framework differ from a generic digital transformation roadmap?
A generic roadmap focuses on tooling and process digitisation. This framework is purpose-built for legal: it diagnoses against a legal-specific six-layer AOS, three transformation phases tuned to legal work, and roles like Legal Engineers. It addresses legal-specific concerns — ethical walls, matter-centricity, audit trails, risk tolerance — and legal-specific use cases like due diligence, closing checklists, and regulatory monitoring that generic frameworks ignore.
How does Phase 2 Differentiation differ from Phase 3 Reinvention?
Phase 2 adjusts service models so a small team competes like a large one, winning on service delivery innovation — new value functions and client relationships emerge within existing structures. Phase 3 rethinks processes from the ground up as an AI-native team: knowledge is productised and deployed at scale through automatic intake, triage, and query response, and entirely new business models emerge. Phase 3 is still new territory.
// Advanced
How do I productise my firm's legal knowledge at scale?
Knowledge productisation is a Phase 3 activity: encode your accumulated knowledge, processes, and standards into the AOS as Skills and automated workflows. Convert static guides, handbooks, and market standards into AI-powered workflows accessible 24/7. Add agentic intake triage that routes queries to the right workflow automatically. This makes expertise scalable and deployable on demand rather than locked in individual lawyers' heads or static documents.
How do I design a human-agent collaboration model with proper oversight?
Integrate oversight throughout, not bolted on at the end. For any long-running task: set the objective, have the agent produce a plan for approval, let it flag uncertainty and ask clarifying questions mid-task, and return completed work to a List for human review. Legal Engineers support deployment, tailoring, and optimisation. The agent surfaces uncertainty and involves human judgment rather than simulating judgment it isn't fit to make.
What new economic models does Agentic Law make possible?
Agentic Law enables managed-service and subscription relationships previously too cost-prohibitive — like continuous regulatory monitoring delivered as a client service, or 24/7 legal guidance via a client portal without linear headcount growth. It unlocks productised legal knowledge, automatic intake and triage, and value functions decoupled from billable hours. The framework asks you to reframe your value proposition around service delivery innovation, not capacity sold by the hour.
How should a 250-lawyer in-house team across multiple jurisdictions apply this?
Diagnose typically as Phase 1 transitioning to Phase 2. Prioritise AOS gaps in the unified data and integrations layer and consistent governance and ethical walls across departments. Design Skills per practice area and jurisdiction, use Lists for shared matter management, and configure Monitors for each regulatory domain. Embed Legal Engineers to drive adoption. A success metric: five legal ops professionals managing AI infrastructure that serves 250 lawyers consistently.
Can a small firm realistically reach Phase 3 Reinvention?
Yes — small firms are often best positioned because they have less legacy structure to unwind. Productise your static guides and market standards into Skills, deploy through a 24/7 client-facing portal with agentic intake triage, and configure Monitors for your clients' regulatory domains. The goal is continuous legal service without linear headcount growth. Legal Engineer capability and change management remain essential even at small scale.