The first generation of generative AI in law has largely followed a familiar model:
lawyer → prompt → AI → answer.
The next generation may look quite different.
Australian startup Mobius is exploring what happens when AI is built into the legal process itself.
Mobius describes its approach as transforming complex, bespoke legal processes into AI-automated production lines using workflows and agents, while allowing organisations to maintain control over their proprietary knowledge and data. (Lander & Rogers)
The distinction is important.
A conventional AI assistant waits for a lawyer to give it individual tasks. An agentic workflow can potentially break a larger piece of work into steps, use different tools or agents to complete them, pass information between those steps, and escalate particular decisions or exceptions to a human.
In other words, the model starts moving from:
AI helps the lawyer do the work
towards:
lawyer designs and supervises a system that does parts of the work.
Mobius was selected for the 2025 Lander & Rogers LawTech Hub, an Australian accelerator that gives legal technology companies access to lawyers, mentors and industry expertise while allowing founders to retain their intellectual property and equity. (Lander & Rogers)
Mobius is still an emerging company, so the interesting thing at this stage is less about judging the eventual product and more about the model it represents.
Agentic AI could ultimately be one of the more consequential developments in legal technology — because it changes the question from “What tasks can AI help lawyers perform?” to “How should we design legal work when AI agents can participate in the workflow?”
That is a much bigger question.
