AI is already part of everyday legal work. Lawyers use it for research, document review, contract comparison, drafting, knowledge management and administration.
That does not make lawyers redundant. It does change what clients expect to pay for, how quickly they expect work back and how much routine effort they are prepared to fund.
Plenty of firms have run a pilot or bought a few licences. The harder question is whether they can use AI repeatedly, safely and profitably on real client work.
Adoption depends on what you count
There is no useful single number for global adoption. Surveys count different things. Some measure whether an individual lawyer has tried ChatGPT. Others ask whether a firm has deployed a legal product across a team. Those are not the same achievement.
The Solicitors Regulation Authority found AI adoption among sole practitioners and small firms in England and Wales at just 14%. Broader surveys of individual lawyers often report much higher use.
The figures should not be treated as a league table. They do point to a practical gap. Individual use tends to run ahead of firm policy, and larger firms usually have more money for specialist products, training and implementation.
The distinction that matters is simple: some firms can support AI as an organisation, while others depend on employees using it quietly and inconsistently.
Clients want proof, not a pitch
Corporate legal teams are adopting AI inside their own organisations. They now expect outside counsel to show that they can use it well too.
In Thomson Reuters' 2026 legal report, 77% of corporate and government clients said AI-enabled improvements to quality were very important or essential. Only 5% said most or all of their providers delivered those improvements.
The demand is reasonable. Clients want quicker, more consistent work without losing a lawyer who is accountable for it.
They are asking direct questions before work begins:
- Does the firm use AI on client matters?
- What information enters the system?
- Is client data retained or used to train a model?
- How are outputs checked?
- Will efficiency be reflected in scope, service or price?
Firms need concrete answers. A vague assurance that they use "secure AI" will not survive procurement or a client's security review.
AI is changing the economics of routine legal work
The billable hour treats time as a rough measure of effort and value. AI makes that logic harder to defend for repeatable work.
When a first-pass review or contract comparison takes less time, a firm has to decide who gets the benefit and how the lawyer's judgement should be priced. Speed matters, but freed-up time is more valuable when the firm uses it to:
- respond to clients sooner;
- analyse more material;
- provide more frequent updates;
- offer fixed-fee services with greater confidence;
- make lower-value matters commercially viable; and
- spend more time on strategy, negotiation and client relationships.
Most firms do not need to build a model. They need to know which tasks cost less with AI, then reflect that saving in the service the client receives.
Small firms can move faster
Large firms have capital, internal data and technology teams. Smaller firms have less room for a bad purchase and fewer people available to test, train and support a new system.
Even so, the SRA found that 82% of small firms and sole practitioners believed technology and innovation would help the sector give clients better value.
Size can also be an advantage. A small practice may be able to choose one workflow, agree on how it should run and train everyone involved without spending months navigating committees.
A 20-person firm that fixes intake and one high-volume workflow can get more from AI than a global firm sitting on hundreds of unused licences.
Budget matters, but follow-through matters more.
Governance has to work on a busy Tuesday
Rules differ by jurisdiction, but the underlying duties are familiar: competence, confidentiality, supervision, candour, client communication and reasonable fees.
In the United States, ABA Formal Opinion 512 applies those duties directly to generative AI. Guidance in Australia and the UK follows the same basic rule. The lawyer remains responsible.
A policy is only useful if someone can follow it in the middle of a working day. Staff need:
- approved tools and prohibited uses;
- clear data-handling rules;
- risk assessment for each workflow;
- compulsory verification of legal content and citations;
- disclosure or consent rules where applicable;
- practical training and incident reporting; and
- someone clearly accountable for oversight.
Good governance removes guesswork. People know which tool to use, what they can put into it and when a human must check the result.
The hard part starts after the pilot
The next step is already visible. Instead of relying on isolated prompts, firms are connecting several tasks into one workflow. A system might summarise an approved meeting, prepare tasks and draft a follow-up from a firm template, with a lawyer reviewing every action.
This is much more useful than a chat window. It also creates more ways for confidential information, a weak instruction or an unchecked output to cause trouble. Security, testing and supervision need to keep pace.
Clients will keep asking what AI changes about the work they receive and the price they pay. Lawyers will expect tools that save more time than they consume. Regulators will still hold the lawyer responsible. None of that is solved by buying licences.
The firms that benefit will be the ones that choose useful workflows, set firm rules and improve both through practice. That work is less exciting than a pilot demo. It is also where the advantage comes from.
LxOS helps law firms choose worthwhile AI use cases, build secure workflows and put sensible controls around them. Talk to us about where to start.