Why ‘agentic’ law is the way forward
Firms need to stop asking what AI can tell us and ask instead what it can do for us, writes Rob Aberdein
For the past few years, the legal profession has been asking what artificial intelligence can tell us. The more important question now is what we can get it to do.
That takes us from generative AI to AI agents and the emerging ‘agentic’ law firm. I believe this represents the next significant change in how legal services are delivered.
Most lawyers still treat AI like a clever junior assistant. Ask it to summarise a document, research an issue or produce a first draft. It responds. Lawyers take the next step.
An AI agent goes further. Give it an objective and, within defined parameters, it can undertake tasks, use different tools and information, determine what needs to happen next and complete work before reporting back or escalating to a human.
Why? Law firms are process-heavy businesses: opening matters, collecting information, identity and anti-money laundering checks, reviewing documents, research, routine correspondence, chasing clients, updating systems, progressing matters, billing. Much of that is necessary. Not all of it needs a human being.
Increasingly, a lawyer will use one agent to review relevant documents, another to research the legal issues, another to identified missing information and a fourth to prepare a first draft. That lawyer has not been replaced, but augmented.
Agents perform tasks. The lawyer remains responsible for strategy, judgment, negotiation, empathy and, critically, client advice. This combination allows individual lawyers to operate with levels of capacity and support previously available only through larger teams.
How? The temptation is to start with the tech. I’d start with the work. Find repetitive tasks consuming significant time. Map how they happen and identify the systems, information and people involved. Ask three questions: what can an agent safely do itself, what should require human approval and what must remain entirely human?
The process is: Identify. Map. ‘Agentise’. Supervise. Scale.
Start small. Give an agent a defined job, controlled access to information and systems it needs, and clear boundaries. Measure the improvement. Then increase its responsibilities or connect agents.
Challenges? There are plenty. Law firms hold extraordinarily sensitive information and operate within professional obligations that don’t disappear because a machine undertakes a task. Confidentiality, privilege, data protection, cybersecurity, accuracy, hallucination, permissions, audit trails, professional indemnity and regulatory compliance all require consideration.
There is also the fundamental question of accountability. The goal is not maximum autonomy. It is appropriate autonomy. The best firms will become exceptionally good at deciding when machines should act independently, when a human should approve something and when technology should stay out of the way.
What is the reward? Efficiency is the obvious answer, but the opportunity is bigger. Agentic working will mean faster transactions, more consistent service, fewer administrative errors, lower costs and substantially greater capacity without equivalent increases in headcount.
It should also make practising law more enjoyable. Remove mechanical work and lawyers can spend more time doing what clients actually value: solving difficult problems, exercising judgment, negotiating, advising and building relationships.
AI will become commonplace across the profession. Simply ‘using AI’ will cease to be a competitive advantage. The real divide will be between firms using AI to make the traditional law firm slightly faster, and those prepared to redesign how legal work gets done.
For law firm leaders, the question is no longer simply what AI can tell your lawyers. It is what you are prepared to let it do.
Rob Aberdein is General Counsel and Group Director of Legal & Strategic Growth, Hoxton Wealth