Nearly seven in ten legal professionals now use AI tools at work. That number more than doubled in a single year. And yet over half of law firms have no AI policy or training programme in place.
That's not a technology problem. It's a people problem. And it's creating real risk.
The Legal AI Skills Gap Is Widening
The gap between AI usage and AI training in law is now a chasm. Lawyers are adopting tools faster than their firms can prepare them.
Thomson Reuters found that 46% of legal professionals report skills gaps in their organisations. Only 22% of law firms have a visible, defined AI strategy. The rest are winging it — and the data shows what happens when they do.
| Metric | Data |
|---|---|
| Legal professionals using AI | 69% in 2026 (doubled in one year) |
| Firms with no AI policy | 53% have none or staff unaware of one |
| Skills gaps reported | 46% of legal professionals |
| Firms with defined AI strategy | Only 22% |
| AI hallucination cases documented | 300+ since mid-2023 |
Firms with a defined AI strategy are 3.5x more likely to experience critical AI benefits and 2x more likely to see revenue growth from AI, compared to those with no strategy. Flying blind isn't just risky — it's expensive.
The Cautionary Tale Everyone Knows (And the Hundreds You Don't)
You've probably heard of Mata v. Avianca — the case where two New York lawyers submitted a brief filled with fake cases invented by ChatGPT. They were fined $5,000 and ordered to personally notify every judge whose name appeared on the fabricated opinions.
That was June 2023. It should've been a wake-up call. Instead, it was the start of an epidemic.
Since then, over 300 cases of AI-generated legal hallucinations have been documented — with more than 200 in 2025 alone. Three separate federal courts sanctioned lawyers in just the first two weeks of August 2025. Cases have appeared across Arizona, Louisiana, Florida, the UK, Australia, Canada, and Israel.
The judge in Mata v. Avianca didn't fault the lawyers for using ChatGPT. The sanctions came because they failed to verify its output. That's the critical distinction: AI itself isn't the problem. Untrained use of AI is the problem.
This isn't about isolated bad actors. It's a systemic training failure happening on an international scale. Every one of those 300+ cases represents a lawyer who used a tool they didn't understand well enough. And that's exactly the gap that structured AI training closes.
What the Bar Associations Are Saying
Professional regulators aren't waiting around. The ABA's Formal Opinion 512 — its first ethics guidance on AI — ties AI use directly to existing professional responsibilities:
- Model Rule 1.1 (Competence): Lawyers must understand the benefits and risks of the technologies they use. This means knowing what AI can hallucinate, when it's reliable, and how to verify its output.
- Model Rule 1.6 (Confidentiality): Entering client data into AI tools without understanding data handling policies could violate confidentiality obligations.
Over 30 U.S. state bars have now released AI-specific guidance, and the CCBE in Europe published its own guide for lawyers using generative AI. The message is consistent: AI competence is no longer aspirational — it's an ethical obligation.
The question for law firms isn't whether to train their people on AI. It's whether they can afford not to.
Where AI Actually Helps in Legal Work
AI in legal practice isn't science fiction. It's already delivering measurable results in specific, high-value use cases.
Contract review is the clearest win. A LawGeex study (via Virtasant) found that AI reviewed NDAs in 26 seconds with 94% accuracy — compared to human lawyers who averaged 92 minutes and 85% accuracy. Legal teams using AI-powered review report saving up to 80% of their time on document analysis.
Beyond contracts, the most common use cases are legal research (74%), document summarisation (74%), and brief drafting (59%). Across all these tasks, lawyers using AI report saving an average of 32.5 working days per year.
The key insight: AI isn't replacing legal judgment. It's eliminating the grunt work that eats into it. Lawyers who know how to use AI for research and review spend more time on the work that actually requires a law degree — strategy, advocacy, and counsel.
But these gains only materialise when people know how to use the tools properly. Without training, AI becomes a liability rather than an asset. That's where the AI governance and prompt engineering skills gap needs to close.
What Effective AI Training for Legal Teams Looks Like
Traditional CLE workshops don't cut it for AI skills. A half-day seminar on "AI in law" might tick a compliance box, but it won't change how anyone works on Monday morning.
The research is clear on what actually works for building lasting AI skills:
Daily practice over one-off events. The forgetting curve is brutal — up to 90% of new information is lost within a week without reinforcement. Short daily sessions with spaced repetition build habits that stick. Six minutes a day beats six hours once a quarter.
Role-specific content over generic overviews. A corporate M&A lawyer and a solo family law practitioner need different AI skills. Generic "intro to AI" training ignores the context that drives real adoption. Contextualised training delivers 40% better comprehension and retention.
Team-based accountability over individual compliance. Learning together creates social pressure to keep going. When a department trains as a group, they develop shared language, shared standards, and shared confidence. That's how you move from "a few people using ChatGPT" to "a firm-wide AI capability."
This is the model kju is built around. Daily sessions tailored to legal industry workflows, covering everything from prompt engineering and AI governance to practical skills like evaluating AI agents and understanding machine learning fundamentals.
The Cost of Waiting
The legal AI market is projected to reach $10.82 billion by 2030, up from $3.11 billion today. 62% of law schools have already integrated AI into their first-year curriculum — meaning junior associates arriving at your firm will expect AI-literate senior partners.
Firms that invest in AI strategy are 2x more likely to see revenue growth and 3.5x more likely to realise critical benefits. Those that don't are training their competitors' future hires.
AI fluency in law isn't something you achieve in a workshop. It's something you build, six minutes at a time.
Frequently Asked Questions
- What AI skills do lawyers need in 2026?
- Lawyers need prompt engineering for legal research, the ability to verify AI-generated citations, understanding of confidentiality risks when using AI tools, and knowledge of their jurisdiction's AI ethics guidelines. These aren't optional extras — the ABA's Formal Opinion 512 makes AI competence part of a lawyer's ethical duty.
- Is AI replacing lawyers?
- No. AI handles routine tasks like document review and contract analysis faster and more accurately than manual methods. But legal judgment, client relationships, courtroom advocacy, and ethical reasoning remain firmly human. The lawyers who thrive will be those who use AI as a tool, not those who compete against it.
- How long does it take to train a legal team on AI?
- Traditional workshops take days but produce poor retention — most knowledge is forgotten within a week. Daily microlearning sessions of 6-10 minutes achieve 80% completion rates and build lasting skills over 8-12 weeks. The key is consistent daily practice, not marathon training sessions.
- What are the risks of using AI without proper training?
- Untrained use of AI in legal practice has led to over 300 documented cases of hallucinated citations, resulting in court sanctions, fines, and reputational damage. Beyond fabricated cases, risks include confidentiality breaches from entering client data into AI tools and failure to meet the ABA's competence requirements.
- Do bar associations require AI training for lawyers?
- Over 30 U.S. state bars have issued AI-specific guidance. The ABA's Formal Opinion 512 ties AI use to existing duties of competence (Rule 1.1) and confidentiality (Rule 1.6). While most jurisdictions don't mandate specific AI training yet, the ethical obligation to understand tools you use effectively requires it.