AI and the future of law are often discussed as though the legal profession is approaching an expiry date. The more useful question is what clients will need from lawyers when much of the preliminary work can be completed with the assistance of artificial intelligence.
AI can help organise documents, generate drafts and explore research questions. It may reduce the time spent on some tasks and change how others are performed. But a client’s problem rarely ends with producing a document. Someone must establish the facts, decide which law applies, weigh the available options and explain the consequences.
That is where the debate needs to begin. The work will change. The need for sound professional judgment remains.
Will AI replace lawyers? Look at the work first
Law is not a single activity called “being a lawyer”. A working day may involve reviewing a contract, interviewing a client, researching a judgment, negotiating a settlement and preparing for court. Each activity requires a different combination of knowledge, experience and judgment.
AI is better suited to some parts of that work than others. The American Bar Association’s Formal Opinion 512 identifies research, contract review, document review and drafting among the activities generative AI can assist with. It also makes clear that using these tools brings professional obligations concerning competence, confidentiality, communication and supervision.
The distinction matters. Software may produce a draft contract, but someone still needs to determine whether its terms suit the transaction. A summary of a judgment may be useful, but it cannot safely replace checking the judgment itself.
The effect on employment and staffing remains uncertain. Some roles may shrink, while others develop around new demands. What seems increasingly difficult to defend is the assumption that legal work will continue unchanged.
The business model will come under pressure
As technology reduces the time needed for certain tasks, clients may begin asking different questions about legal fees.
Why did a routine document take so long? What required specialist attention? What did the lawyer contribute beyond the first draft?
These are reasonable questions. They do not diminish the value of legal expertise. They require firms to explain that value more clearly.
A document produced quickly may still need substantial review. The facts may be incomplete, the instructions unclear or the consequences unusually serious. Time saved in drafting does not necessarily mean that the matter is simple.
Nevertheless, efficiency should have a visible purpose. It might allow a firm to respond sooner, offer a more accessible service or devote more attention to difficult aspects of a matter.
The ABA’s guidance provides a concrete example. Under its approach, a lawyer billing by the hour must charge for the actual time spent, including necessary review, rather than the time the task might have taken without AI.
This is American professional guidance, rather than a statement of South African billing requirements. It nevertheless illustrates a question that firms everywhere will need to consider: how should technological efficiency affect what clients pay?
Better tools require better working methods
Buying an AI subscription is relatively straightforward. Integrating it responsibly into a legal practice requires more thought.
A firm needs to decide which tasks are suitable, what information may be entered and who will review the output. It also needs to recognise when using AI adds an unnecessary step.
A sensible starting point is a clearly defined task with a result that can be checked. The lawyer should know what a satisfactory answer would require and be able to identify material omissions.
Useful questions include:
- What are we asking the tool to do?
- Does it have the information needed for that task?
- Can we independently verify its output?
- What could happen if the result is wrong?
- Who is responsible for approving its use?
These questions keep attention on the quality of the work. The objective is to improve the service, rather than simply increase the amount of content a firm can produce.
Junior lawyers still need to learn the foundations
The training question deserves particular attention.
Junior lawyers have traditionally developed their skills through research, document review and first drafts. Some of that work is repetitive, but it also teaches them how to recognise an issue, distinguish relevant facts and understand the structure of a legal argument.
If AI performs the preliminary work, firms must consider how juniors will acquire those foundations.
It is difficult to assess a machine’s answer without knowing enough to question it. A convincing draft may conceal a missing qualification. A plausible citation may refer to an authority that does not exist. A correct general statement may be unsuitable for the client’s circumstances.
Training should therefore give juniors opportunities to:
- Conduct research independently and compare it with AI output.
- Explain why an authority supports a particular conclusion.
- Identify what a draft leaves out.
- Interview clients and test assumptions.
- Discuss their reasoning with experienced practitioners.
AI can form part of that learning process. It should not become a shortcut around understanding the work.
Responsibility cannot be delegated to software
Clients approach lawyers because something important is at stake. It may be their home, livelihood, business, family or freedom.
The professional relationship requires more than an answer. It requires attention to instructions, confidentiality, communication and the consequences of acting on advice.
AI does not assume those responsibilities on the lawyer’s behalf.
The judiciary of England and Wales addresses this issue in its October 2025 guidance. It warns about incorrect or misleading AI output, risks arising from biased training data and the disclosure of private information. It also reinforces judicial office holders’ personal responsibility for material produced in their names.
For a South African practitioner, verification must include checking that the authority belongs to the correct jurisdiction and remains applicable. A fluent explanation of foreign law can be misleading when presented as an answer to a local legal problem.
Confidentiality also requires attention before information is entered into a tool. The ABA’s opinion advises lawyers to assess disclosure risks and understand the relevant terms and privacy arrangements.
Clients will still need someone to listen
A client does not always arrive with a clearly defined legal question. They arrive with an account of what happened, what worries them and what they hope to achieve.
The important fact may emerge only after careful questioning. A client asking about litigation may primarily want a business relationship preserved. Someone demanding an immediate response may not yet understand the costs or risks involved.
Legal judgment includes recognising these differences.
Technology may assist with preparing options or organising information, but the lawyer still needs to establish what matters to the client. Advice becomes useful when it connects the law with the person’s actual circumstances.
Clear communication will therefore remain an important part of professional value. Clients need to understand what can be done, what remains uncertain and which decisions require their instructions.
Access to legal support could improve
AI could help people understand unfamiliar terminology, prepare for a consultation and identify the kind of assistance they need. It could also help smaller practices manage routine work more efficiently.
These possibilities deserve attention, particularly where cost and uncertainty discourage people from seeking help.
But better access to information does not automatically mean better access to justice. A person may misunderstand an answer, overlook a deadline or rely on guidance that does not apply to their circumstances.
Legal platforms should explain the purpose and limits of their services. General education, preliminary guidance and professional representation serve different needs.
The value lies in helping people reach an appropriate next step, including knowing when they need advice from an attorney.
Trust will become more important
When polished legal writing is easy to produce, presentation alone becomes a weaker indication of reliability.
Readers will need to ask who wrote the material, what supports it and whether it applies to their situation. Named authors, relevant expertise, accessible sources and accurate dates help them make that assessment.
For firms, trust also develops through ordinary service: returning a call, explaining a delay, admitting uncertainty and following through on an undertaking.
AI may assist with those processes, but the firm remains responsible for the experience it provides.
Legal publications and professional platforms have a useful role here. They can help readers distinguish informed explanation from unsupported assertion and connect public questions with identifiable attorneys.
The future of law will reward judgment
No one can confidently predict every effect AI will have on the legal profession. Its capabilities, costs and limitations will continue to influence how firms use it.
The practical response is to examine where it improves work, where it introduces risk and where professional involvement remains essential.
Lawyers will need to understand the tools they use without treating them as unquestionable authorities. Firms will need to preserve training while improving efficiency. Clients will need clear explanations of the service they are receiving.
The profession’s value will increasingly depend on what lawyers do with the information available to them: the questions they ask, the advice they give and the responsibility they accept.
AI may change how legal work begins. Lawyers will still have to decide how it should proceed.
