When AI handles much of the reading, extracting and first drafting, the shape of legal work changes. The operating model, meaning who does what, with which tools and at what cost, needs redesigning rather than patching.
What changes
- Work mix: routine work (NDAs, standard contracts, simple questions) moves to self-service and AI-assisted processes; lawyers spend more time on judgment, negotiation and advising.
- Roles: new roles appear, such as legal engineers, legal ops and AI product owners for legal; paralegal and junior work shifts toward review, verification and process design.
- Intake and self-service: one front door for the business, with AI triage, playbook-driven self-service for low-risk requests and routing to the right lawyer for the rest.
- Knowledge: templates, playbooks and past advice become a managed asset, because AI is only as good as the material it draws on.
- Sourcing: the split between in-house, outside counsel and alternative legal service providers is reconsidered task by task.
Design principles
- Start from the work, not the tool: map the matter types and volumes first.
- Tier by risk: self-service for low risk, AI-assisted lawyer work for medium, lawyer-led for high.
- Keep humans accountable: every tier has a named owner for the outcome.
- Measure: cycle time, cost, quality and business satisfaction by tier.
- Train for new skills: verification, prompt design, process thinking and data literacy.
Regulators are adapting too
In England and Wales, the Solicitors Regulation Authority authorized Garfield.Law in May 2025 as what it described as the first purely AI-based firm to provide regulated legal services, after satisfying itself about quality checks, confidentiality and conflict safeguards, with named solicitors remaining accountable.