Defining Real Agentic AI in Legal Tech Beyond Workflow Hype
Jake Jones of Flank clarifies what truly constitutes agentic AI in legal tech, distinguishing it from overhyped workflow tools and offering a framework for evaluation.
By Jake Jones, Flank
Legal tech is increasingly mislabeling basic workflow tools as 'agentic AI,' creating confusion and slowing industry progress. True agentic AI must autonomously pursue goals, adapt to obstacles, and operate without constant human oversight. This article outlines the key characteristics of genuine agentic systems and exposes common misconceptions.
Key Features of Agentic AI
- Autonomy: Operates unattended, re-plans when faced with obstacles, and dynamically selects tools.
- Policy Compliance: Enforces rules and risk tolerances with auditable logs.
- Channel-Native: Works via existing platforms like email, Slack, or Teams, not proprietary UIs.
What Agentic AI Is Not
- Workflow Wrappers: Rigid processes with LLM prompts that fail under real-world variability.
- Integration Theatre: Systems requiring manual tool selection despite claiming autonomy.
- Play-Acting Copilots: Tools that draft content but cannot execute end-to-end tasks.
Evaluating Agentic Claims
Buyers should demand metrics like:
- Unattended Completion Rate (UCR): Percentage of tasks completed without human intervention.
- Obstacle Recovery Rate (ORR): Ability to resolve blockers autonomously.
- Mean Time to Human (MTTH): Average runtime before human input is needed.
- Policy Breach Rate (PBR): Instances of out-of-policy actions per 1,000 runs.
Architecture Matters
True agentic systems require:
- Planner/Controller: For goal-setting and adaptive planning.
- Policy Engine: To enforce constraints and thresholds.
- Audit Layer: For immutable logs and reproducibility.
A Concrete Example: NDA Execution
A real agent can:
- Parse intake from email/Slack, classify risk, and select the correct template.
- Draft, negotiate within authority, and escalate only when necessary.
- Update systems like CLM/CRM and notify stakeholders autonomously.
The Paradigm Shift
Agentic AI represents a new class of systems—intelligent, autonomous, and capable of acting across your tech stack. Vendors must stop rebranding workflows and instead build or market products honestly.
About the Author: Jake Jones is co-founder of Flank, a legal tech company developing autonomous agents for routine legal tasks.
Upcoming Events:
- Legal Innovators UK (Nov 4-6, 2025)
- Legal Innovators New York (Nov 19-20, 2025)
This article was originally published on Artificial Lawyer.
Related News
Agentic AI Hype Faces Reality Check Amid Privacy and Performance Concerns
The resurgence of Agentic AI promises automation and efficiency, but privacy risks and high failure rates reveal a gap between hype and reality.
Agentic AI in Legal Workflows Key Takeaways from ILTACON 2025
Agentic AI offers autonomous legal workflow potential but requires proper implementation with task delegation data context and human oversight as noted by an ILTACON panel
About the Author

Dr. Lisa Kim
AI Ethics Researcher
Leading expert in AI ethics and responsible AI development with 13 years of research experience. Former member of Microsoft AI Ethics Committee, now provides consulting for multiple international AI governance organizations. Regularly contributes AI ethics articles to top-tier journals like Nature and Science.