Preparing Trainee Advocates for AI-Enabled Legal Practice

A pupil at a Kenyan law firm submits a legal opinion containing a persuasive analysis of a judicial decision, complete with a citation that appears entirely credible. The reasoning is coherent, the language is professional, and the opinion seems ready for review. However, when the supervising advocate attempts to retrieve the judgment from Kenya Law, no such decision can be found. Further questions reveal that the pupil relied on an AI-generated response without independently verifying the authority.

What initially appeared to be competent legal research turns out to be a fundamental failure of professional judgment. The incident illustrates a growing challenge for Kenya's legal profession: how do we prepare aspiring advocates to practise competently in an environment where artificial intelligence can generate legal opinions, draft pleadings, summarise judgments and suggest legal arguments within seconds?

This question is particularly relevant to Kenya's pupillage system, which forms an essential component of the Advocates Training Programme administered through the Kenya School of Law. Pupillage is intended to bridge the gap between academic legal education and professional practice by exposing aspiring advocates to the responsibilities, ethical obligations and practical demands of legal work under the supervision of experienced practitioners.

As AI becomes increasingly integrated into legal practice, that supervisory responsibility must evolve. Pupils need to learn not only how to use emerging technologies, but also how to interrogate their outputs, recognise their limitations and exercise independent legal judgment.

Drawing on Nelson Nkari's training at the Kenya School of Law's Virtual Pupil Masters' Workshop on 10 April 2025, this article explores how Kenyan law firms and supervising advocates can incorporate AI into pupillage without compromising the professional standards expected of advocates practising in Kenya.

Teach pupils what the tool is doing

The responsible use of AI in Kenyan legal practice begins with understanding the technology itself. Large language models, which power many contemporary AI assistants, generate text by drawing on patterns learned during training and information provided to them. Their ability to produce coherent and convincing responses makes them valuable tools for summarising judgments, drafting legal documents, reviewing contracts and suggesting possible lines of legal research.

However, these capabilities come with important limitations. AI-generated responses may contain factual inaccuracies, omit relevant legal qualifications, misinterpret statutory provisions or refer to cases that do not exist. More concerning is the fact that such errors may be presented in language that appears authoritative and professionally sound.

For a pupil researching Kenyan law, the consequences can be significant. An AI tool may confidently attribute a legal principle to a decision of the High Court, Court of Appeal or Supreme Court of Kenya without accurately representing the court's reasoning. It may also overlook subsequent decisions, legislative amendments or procedural developments that materially affect the applicable law.

Pupils must therefore appreciate that a convincing legal explanation is not necessarily a correct one. The reliability of any legal proposition depends on the authority supporting it, the accuracy with which that authority has been interpreted and its relevance to the particular legal issue.

This understanding should extend beyond generative AI. Different legal technologies perform different functions and consequently present different risks. Document-management systems may use algorithms to classify or retrieve information, while machine-learning applications may analyse historical data to identify patterns and generate predictions. Generative AI tools, by contrast, produce new content in response to user instructions.

These distinctions become particularly important when considering technologies such as litigation analytics. A platform that purports to predict litigation outcomes based on previous court decisions must be evaluated in light of the quality, representativeness and completeness of its underlying data. In Kenya, where the availability and publication of judicial decisions may vary across courts and time periods, such limitations are especially relevant.

A pupil should be capable of explaining these limitations before relying on an AI-generated prediction or presenting it to a supervising advocate or client. The objective is not necessarily to turn aspiring advocates into technology specialists, but to equip them with sufficient technological literacy to use these tools critically and responsibly.

Keep legal reasoning visible

At the heart of Kenya's pupillage framework is the development of professional competence through supervised practical experience. The purpose of pupillage is not simply to familiarise aspiring advocates with the daily operations of a law firm, but to cultivate the analytical ability, ethical awareness and professional judgment required of an advocate of the High Court of Kenya.

Artificial intelligence should complement this learning process rather than undermine it.

Traditionally, a pupil assigned a legal research question would be expected to identify the relevant statutory provisions, examine judicial decisions, assess competing interpretations and formulate a reasoned legal opinion. The supervising advocate could then evaluate the pupil's understanding by questioning the sources relied upon and the reasoning supporting the conclusion.

The growing accessibility of AI-generated legal content creates a risk that pupils may produce sophisticated-looking work without undertaking the intellectual process necessary to understand it. A pupil who submits an AI-generated legal opinion may be unable to explain the relevant statutory framework, distinguish an adverse authority or defend the legal conclusion when challenged.

This possibility makes it increasingly important for pupil masters to evaluate the reasoning behind submitted work, rather than relying solely on the quality of the final document.

One practical approach is to require pupils to undertake selected assignments independently before introducing AI assistance. For example, a pupil preparing a commercial agreement should first demonstrate an understanding of the purpose of its provisions, the allocation of contractual risks and the legal consequences of particular drafting choices.

Once the pupil has completed an initial draft, an approved AI tool may be used to generate an alternative version. Comparing the two documents provides an opportunity to examine differences in wording, identify omitted safeguards and discuss whether particular provisions adequately protect the client's interests under Kenyan law.

The same approach can be adopted in litigation. A pupil may initially prepare submissions based on the Civil Procedure Act, the Civil Procedure Rules and relevant Kenyan judicial authorities before comparing the work with an AI-assisted draft. The resulting discussion should focus on the accuracy of the legal arguments, the applicability of the authorities and the procedural requirements governing the matter.

Through these exercises, pupils learn that AI is a tool for assisting legal work rather than a substitute for legal reasoning.

Efficiency should also be assessed realistically. Although AI may generate a draft within seconds, the time required to verify citations, correct inaccuracies and ensure compliance with Kenyan law may substantially reduce the apparent time saved.

The real measure of successful AI adoption is therefore not the speed at which a document is produced, but whether the final work is accurate, properly supported and capable of being defended by the advocate responsible for it.

Design research exercises around primary material

Legal research is one of the areas in which AI offers immediate practical benefits to Kenyan law firms. It is also an area where uncritical reliance on generated content can have particularly serious consequences.

Kenyan advocates routinely work with legislation, subsidiary legislation, reported and unreported judicial decisions, procedural rules and other authoritative materials. The ability to identify, interpret and apply these sources remains fundamental to competent legal practice.

Pupillage training should therefore reinforce the importance of primary legal authorities, particularly when AI tools are introduced into research workflows.

A useful exercise would involve assigning a pupil a publicly available Kenyan judgment and requiring them to identify the legal issues, summarise the court's reasoning and explain the principle established by the decision. The pupil should also locate the specific passages supporting the analysis and consider whether subsequent decisions or legislative developments have affected the judgment's relevance.

The pupil may then use an approved AI tool to summarise the same decision and compare its output with the original judgment.

Such an exercise can reveal how AI systems sometimes omit important qualifications, mischaracterise judicial reasoning or present broader legal conclusions than the underlying authority supports.

Kenya Law, operated by the National Council for Law Reporting, provides an important starting point for locating Kenyan legislation and judicial decisions. Pupils should become proficient in using this resource and other appropriate primary sources to verify authorities before relying on them in legal opinions, pleadings or submissions.

Verification, however, involves more than confirming that a cited case exists. A pupil must understand the court that decided it, the context in which the issue arose, the reasoning supporting the decision and whether the authority is binding, persuasive, distinguishable or otherwise applicable to the matter under consideration.

For instance, a decision of the High Court may provide useful guidance on a legal question, but its precedential significance must be assessed within Kenya's judicial hierarchy and in light of any relevant decisions of the Court of Appeal or Supreme Court.

Research exercises should also introduce pupils to the risks associated with relying on foreign authorities generated by AI systems. Many general-purpose AI tools draw extensively on legal materials from jurisdictions such as the United States and the United Kingdom. While comparative jurisprudence may sometimes be persuasive, pupils must appreciate that foreign legal principles cannot automatically be applied to Kenyan disputes without considering the relevant constitutional, statutory and jurisprudential context.

These principles can be reinforced through practical document-review exercises. Pupils may be given public or synthetic contracts and asked to use AI to identify obligations, liability provisions, termination rights and other material terms. Their findings can then be compared with a manual review to establish whether the system overlooked exceptions or misinterpreted contractual language.

The emphasis should remain on the legal implications of any errors identified. An incorrectly interpreted indemnity clause, for example, may fundamentally change the advice given to a client. Connecting technological errors to their legal consequences helps pupils develop the professional scepticism necessary for responsible practice.

Protect confidential information during training

Confidentiality is a foundational obligation of legal practice in Kenya. Advocates routinely handle sensitive personal information, privileged communications, commercial records, financial information and documents relating to ongoing disputes. The introduction of AI technologies into law firms creates additional considerations concerning how such information is processed, stored and potentially disclosed.

A publicly accessible AI platform should not automatically be regarded as an appropriate destination for client information merely because it is convenient or capable of performing a particular task.

Before authorising its use, a law firm should establish how the service handles information submitted by users, whether that information may be retained or used for model training, who may access it and where it is processed or stored.

These concerns are particularly significant where pupils are experimenting with AI tools as part of their training. Without appropriate guidance, a pupil may upload a client's agreement, witness statement or confidential legal correspondence into a public AI service without appreciating the implications.

Even removing obvious identifiers may not adequately protect confidential information. The remaining details of a transaction, dispute or factual circumstances may still make the client identifiable.

Kenya's Data Protection Act, 2019, and its subsidiary regulations provide an important framework for assessing the processing of personal data through AI services. Depending on the circumstances, the firm may need to consider the applicable lawful basis and purpose of processing, the responsibilities of data controllers and processors, appropriate security safeguards and requirements governing transfers of personal data outside Kenya.

The constitutional right to privacy under Article 31 of the Constitution of Kenya also provides an important foundation for understanding the protection of personal information. These data-protection considerations exist alongside the professional duties of confidentiality and obligations relating to legal professional privilege.

Pupillage training should translate these principles into practical workplace procedures. Initial AI exercises should ordinarily use public judgments, publicly available legislation and synthetic documents that do not contain actual client information.

Where a proposed exercise involves real client material, the pupil should understand that approval must first be obtained through the firm's established procedures and that the proposed use must be assessed against the relevant legal, contractual and professional obligations.

Supervisors can also incorporate practical exercises involving the review of an AI vendor's privacy policy, contractual terms, data-retention settings and security controls. Such exercises are particularly valuable because they introduce pupils to the growing intersection between legal practice, technology procurement and data-protection compliance.

Equally important is ensuring that pupils understand what to do when a mistake occurs. Where confidential information has inadvertently been submitted to an unapproved platform, the firm should have a clear reporting and response process. Prompt disclosure enables the firm to assess the circumstances, evaluate the potential consequences and determine what corrective action may be required.

Give supervisors a workable policy

Responsible AI adoption within Kenyan law firms requires more than individual awareness. Firms need clear institutional policies that define how AI may be used, who may authorise its use and what safeguards must be followed.

This is particularly important during pupillage, when aspiring advocates are still developing professional habits and learning the standards expected within legal practice.

A practical AI policy should identify approved tools, specify permitted uses and establish restrictions on the information that may be entered into those systems. It should also make clear that AI-assisted work remains subject to appropriate human review before being communicated to clients, filed in court or relied upon in professional decision-making.

For example, a firm may permit pupils to use an approved AI tool to summarise publicly available judgments while prohibiting the submission of confidential client documents into unapproved services. It may also require all AI-generated legal authorities to be independently verified and all substantive legal conclusions to be reviewed by the supervising advocate.

The effectiveness of such a policy depends on its integration into everyday supervision.

Pupil masters should develop sufficient familiarity with the tools being used to ask meaningful questions about their outputs. When reviewing an AI-assisted legal opinion, the supervisor should be able to establish which primary authorities support the conclusions, whether the relevant documents were considered in full and whether the analysis accurately reflects current Kenyan law.

Supervisors should also encourage pupils to disclose when and how AI was used in preparing assignments. Transparency makes it easier to evaluate both the legal work and the pupil's understanding of the technology.

Importantly, supervision should not become an exercise in detecting or punishing AI use. Where a firm permits appropriately controlled use, the objective should be to cultivate professional responsibility, critical evaluation and sound judgment.

An effective policy should also address incidents involving inappropriate use, particularly where confidential information or sensitive material may have been exposed. Pupils should know whom to notify, what information to provide and how the firm will assess the situation.

Clear procedures and consistent supervision are more likely to promote responsible behaviour than informal restrictions that leave pupils uncertain about what is permitted.

Evaluate tools before expanding their role

The growing availability of AI products marketed to legal professionals can create pressure on law firms to adopt new technologies quickly. However, a system's ability to produce impressive demonstrations does not necessarily establish its suitability for professional legal work in Kenya.

Before introducing an AI platform into pupillage or wider firm operations, legal practitioners should assess its performance against the actual tasks for which it is intended.

A firm might begin with a controlled pilot involving publicly available Kenyan judgments, legislation or synthetic agreements. The objective would be to determine whether the tool produces accurate and useful outputs, how frequently errors arise and how much additional work is required to verify its results.

Particular attention should be paid to whether the system reliably handles Kenyan legal materials. A general-purpose AI assistant may perform well when discussing broadly recognised legal concepts but be less dependable when interpreting specific provisions of Kenyan legislation, analysing recent local judgments or addressing procedural requirements.

This distinction matters because legal accuracy is inherently jurisdiction-specific.

An AI-generated answer that correctly describes a principle of English or American law may nevertheless be inappropriate for a Kenyan legal opinion. Similarly, an inaccurate summary of a Kenyan judicial decision may create substantial risks if relied upon in litigation or client advice.

Risk assessment should therefore consider both the nature of the information processed and the consequences of an incorrect output reaching a client, regulator or court.

Even where an AI tool processes only publicly available information, the resulting work may still be high risk if it concerns legal advice, pleadings or submissions. The absence of confidential data does not eliminate the need for rigorous verification.

Firms should also recognise that AI systems evolve. Changes to a model, product integration, hosting arrangement or data-processing policy may affect whether a previously approved service remains appropriate.

Maintaining a record of the tools evaluated, their configurations, approved uses and identified limitations provides a practical foundation for periodic reassessment.

Involving pupils in these evaluations can itself be educational. It exposes aspiring advocates to technology-risk assessment and encourages them to approach AI adoption as a professional decision requiring evidence and judgment, rather than simply a matter of convenience.

Prepare pupils for a wider range of work

The integration of AI into legal practice is occurring alongside broader technological changes affecting Kenya's legal services sector. The digitisation of court processes, increasing reliance on electronic evidence, expanding use of digital financial services and growing importance of data protection and cybersecurity are changing the nature of legal work.

These developments create opportunities for advocates whose professional competence extends beyond traditional drafting and litigation to include an understanding of the technologies shaping modern commercial and public-sector activity.

Kenya's expanding digital economy has already created demand for legal services relating to technology contracts, fintech regulation, intellectual property, cybersecurity, personal-data protection and digital governance. AI adds further dimensions, particularly in relation to automated decision-making, accountability, technology procurement and the governance of systems used by businesses and public institutions.

Pupillage should provide aspiring advocates with opportunities to appreciate these developments and understand how established legal principles apply to emerging technologies.

This does not mean that every pupil must become a programmer or specialist in artificial intelligence. Rather, pupils should develop sufficient technological literacy to engage meaningfully with technical professionals, understand the practical context of legal problems and translate technological risks into appropriate legal advice.

For instance, an advocate reviewing an agreement for the procurement of an AI-enabled service may need to understand how the system processes information, allocates responsibility between the customer and provider, protects confidential data and addresses inaccurate or harmful outputs.

Exposure to discussions involving software developers, cybersecurity professionals and information-governance teams can help pupils develop the interdisciplinary understanding required to advise effectively on such matters.

At the same time, the increasing use of technology should not diminish the importance of traditional professional competencies.

Kenyan advocates must still be capable of interviewing clients, assessing evidence, interpreting legislation, conducting legal research, preparing pleadings, negotiating agreements and presenting persuasive arguments before courts and tribunals. These responsibilities require contextual understanding, ethical awareness and independent professional judgment.

AI may assist with preparation, but it cannot relieve an advocate of responsibility for the work undertaken or the advice provided.

Continuing professional development should therefore be approached as an ongoing process rather than a one-off introduction to new technology. Law firms can incorporate practical AI-related learning into ordinary pupillage supervision, internal training sessions and professional discussions.

Reviewing changes to a technology vendor's contractual terms, examining an AI-generated legal opinion for inaccuracies or testing a new research tool against Kenyan authorities can provide valuable learning experiences.

One particularly effective exercise is to require a pupil to identify an AI-generated recommendation or conclusion that they chose not to adopt and explain the legal reasoning behind that decision. The ability to recognise when an apparently convincing answer is incorrect, incomplete or inappropriate is an important indicator of developing professional judgment.

Ultimately, Kenya's legal profession should approach AI integration into pupillage as an opportunity to strengthen, rather than weaken, the quality of practical legal training.

The objective should not be to produce advocates who can generate documents more quickly, but to develop professionals who understand the law, can evaluate technological outputs critically and are capable of defending their conclusions through sound legal reasoning.

For pupil masters, this requires a deliberate balance between embracing technological innovation and preserving the essential disciplines of legal practice. For pupils, it requires recognising that AI proficiency is valuable only when accompanied by legal knowledge, professional ethics and independent judgment.

As Kenyan law firms continue to adopt AI and other legal technologies, the advocates best prepared for the future will be those who can combine technological competence with a firm grounding in Kenyan law and the professional responsibilities of an advocate.

Artificial intelligence may change how legal work is performed, but it does not change the advocate's responsibility to ensure that the work is accurate, ethical and legally sound.

Our technology-law practice advises Kenyan law firms and other organisations on AI governance, responsible technology adoption, data protection and professional training. Contact our firm to discuss an AI training programme tailored to the needs of your legal workplace and its professional and regulatory obligations.

Preparing Kenya's Next Generation of Advocates for AI-Enabled Legal Practice