Most conversations about AI and law begin with the same question: will AI replace lawyers?
That question matters, but I think it misses the change already happening.
AI is changing what organisations expect lawyers to do. It is changing how legal work enters a team, who handles it and where human judgment is needed.
In a recent Financial Times report on innovative in-house legal teams, Repsol described the shift in one line:
From “I do the work” to “I design how work happens”.
That line stayed with me.
The lawyer is no longer responsible only for answering the question or producing the document. Increasingly, the lawyer must also help design the process around the work.
A good legal answer still needs a good system
Legal education teaches us how to find the law, interpret it and apply it to a set of facts. Those skills remain essential.
But an in-house lawyer works inside an organisation. The answer is affected by how the request arrives, whether the right facts were collected, who can approve the decision and whether the advice reaches the right person in time.
A legally correct answer delivered too late, to the wrong person or through a process nobody understands may still be poor legal service.
This is what makes the examples in the FT report interesting. Microsoft created a central “front door” for legal requests. AI sorts routine questions, while lawyers step in when judgment is needed. Remote trains selected lawyers in software engineering. Repsol trains its lawyers in process design, decision analytics, AI prompting and project management.
The point is not that every lawyer must become a programmer. The point is that legal knowledge now has to connect with how work actually moves through an organisation.
Someone still has to decide where judgment belongs
Generating a contract or sorting a request is one thing. Deciding when a human lawyer must intervene is much harder.
Someone has to decide which matters are routine, which facts make a case unusual, what level of risk requires escalation and who is responsible when the system gets it wrong.
Those are not simply technology decisions. They are legal and ethical decisions.
A good system does not remove judgment. It puts judgment where it matters most.
Legal knowledge is not becoming optional
There is a risk that all this talk about legal engineers and AI makes traditional legal knowledge sound less important.
It is not.
If the legal analysis is weak, AI simply allows weak analysis to travel faster. If the escalation rules are badly designed, the system may produce consistency without producing a fair result.
The lawyer designing the process must understand the law well enough to know what can be automated and what must remain with a person.
As someone progressing towards legal qualification after working across psychology, education, safeguarding and project delivery, this part resonates with me. Legal problems rarely exist by themselves. They sit inside organisations, deadlines, incentives, relationships and human behaviour.
Understanding those systems does not make the work less legal. It can make the legal judgment more useful.
The profession is becoming more interdisciplinary
Repsol gave its legal team of almost 200 people around 100 hours a year for programming and AI training. The team estimates that its new practices saved more than 22,000 hours in one year. It also reduced external legal spending by more than €360,000 on one matter.
That is not a small experiment on the side. It is a different way of organising legal work.
Future lawyers will not all need the same technical skills. Some will build tools. Others will focus on risk, data, project delivery or complex advice. But they will all need to ask better questions:
- What problem is this system solving?
- What information is it relying on?
- Where can error or bias enter?
- Which decisions still require a person?
- Who remains responsible?
Leaders have to be willing to learn
Prosus made senior leaders sit with junior colleagues and build a working AI tool. Its group general counsel admitted that the experience was intimidating because he did not have a coding background.
I think that matters.
Many organisations announce an AI strategy while continuing to organise work exactly as before. Junior staff are told to experiment, while senior leaders remain distant from the tools and the difficult questions they create.
Seniority should not excuse someone from learning. If leaders want people to change how they work, they should be willing to experience the discomfort of learning something new themselves.
Efficiency cannot simply mean more work
Saving 22,000 hours sounds impressive. But what happens to those hours?
If every saved hour is immediately replaced with another task, the result may simply be a faster and more exhausted legal team.
The real value is in creating space for earlier advice, better analysis, clearer communication and the difficult matters where judgment makes a difference.
AI should not only help lawyers do more work. It should help them spend more time on the work that matters most.
We also have to think about junior lawyers
Junior lawyers develop judgment by researching, drafting, reviewing documents, checking details and learning from correction.
Those are some of the first tasks organisations want to automate.
If firms remove the work through which people learned, they must build a new route to competence. Otherwise, they may gain efficiency today and create a shortage of experienced lawyers tomorrow.
Junior lawyers need supervised responsibility, exposure to the reasoning behind decisions and opportunities to test AI outputs rather than simply accept them.
Law firms will have to show how they create value
Revolut is already using data on cost, quality and turnaround time when deciding which external law firms to instruct. It also asks firms how they use technology and what benefit that use produces.
This changes the question from “How many hours did this take?” to “Why is the work being done this way?”
Simply having an AI licence will not be enough. Firms will need to show that technology improves the work without weakening confidentiality, supervision or responsibility.
The lawyer of the future designs responsibility
The future lawyer will still need to understand the law, exercise judgment, negotiate and give clear advice.
But the role is getting wider.
Lawyers will also help decide how requests enter a team, what can be automated, where people intervene, how quality is checked and who answers when something goes wrong.
The question will not only be whether the work was done correctly. It will also be whether the system was designed to produce good work consistently.