Thibaut Caoudal’s Post

Unpopular opinion: Better AI tools won't fix broken legal processes. I keep seeing this pattern. Organizations rush to adopt the latest LLM. They expect transformation. Instead? Confusion. Because here's the truth: AI only makes sense when the underlying workflow does. You can't automate chaos. You need clean data first. Predictable workflows. Clear boundaries. Real ownership. Most APAC legal teams I work with aren't limited by AI capability. They're limited by operational foundations. The teams winning right now? They're not chasing moonshots. They're building systems that make attorney knowledge usable at scale. They understand their edge cases. They have clear requirements. They know exactly what problems they're solving. Then-and only then-they bring in AI. And when they do? The practical wins show up fast. 20-30% efficiency gains in contract review. Faster turnaround times. Business teams actually using legal tools. This isn't about replacing lawyers. It's about repositioning them. Freeing them from cognitive overload so they can focus on judgment, strategy, and real value creation. The industry noise-who's ahead, who's behind, which benchmark matters-is mostly irrelevant to getting work done. What matters? Steady progress. Clear workflows. Collaborative execution. The organizations that master this will look back in a year and barely recognize their operations. Are you building foundations or chasing features? ➡️ Follow ➡️ Connect ➡️ Comment ➡️ Refer ➡️ DM to inquire further

Most law firms don’t fail with AI because of hallucinations. They fail because no one redesigns the process around it. I’ve been testing legal AI workflows hands-on for research, drafting, review, and operational tasks. The pattern is consistent: • AI speeds up one isolated step • Everything downstream stays unchanged • Risk quietly accumulates Example: Automating legal research without redefining review standards doesn’t save time. It compresses timelines while expanding the liability surface. The uncomfortable truth: AI is not a productivity tool. It’s a process amplifier. Weak processes get riskier. Strong processes get faster. Over the next few weeks, I’ll be breaking down: • where legal AI actually works • where it predictably breaks • and how firms can adopt it without operational chaos If you’re already experimenting with AI in legal work, which part of the workflow feels most fragile right now?

A sobering (and deliberately provocative) vignette about where senior lawyers think AI is heading. Over martinis in Soho, a highly respected English barrister (who remains anonymous) describes running a real appellate judgment through a cutting-edge AI. What took him a day and a half, the system produced in seconds — and, in his view, better. Not just competent, but KC-level drafting for pennies. His conclusion is stark: legal AI won’t just assist lawyers, it will replace large parts of the profession. First the grunt work, then drafting, then complex reasoning. Process-heavy areas like probate, conveyancing and advisory work are, he says, especially vulnerable. Even advocacy and judging may not be immune once economics trump sentimentality. Concerns about hallucinations, ethics, and the need for a human face in court are waved away as “temporary bugs”. The real driver, he argues, is cost: once AI can outperform humans faster and cheaper, resistance won’t last. Attempts to ban or restrict it will fail. He’s equally scathing about professional denial. Many lawyers reassure themselves that AI is “just a tool”. He thinks that’s delusion — and that professional arrogance will make the reckoning harder. The psychological and economic shock, he predicts, will ripple well beyond law. His bleakest advice? Don’t encourage young people into legal careers built on debt and assumptions of permanence. In his words: don’t train for a job that may not exist in ten years — or less. Whether one agrees or not, the piece forces an uncomfortable question: are we preparing the profession for augmentation, or for displacement — and do we really understand the difference yet?

Calling out this BS. I’ve been in the legal tech space for over 10 years, and every few months the same headline comes back claiming AI will replace lawyers.  It sounds dramatic, but it shows a poor understanding of how legal work actually functions in the real world. You don’t engage a lawyer to write a document. You engage a lawyer to hold the responsibility for your case. When something goes wrong, it’s not the output that gets questioned, it’s the judgment behind it. Accountability, context, and risk don’t disappear because drafting got faster. We’ve been building AI and technology in law for a decade now, and I’m still 100% sure this isn’t happening. Lawyers are some of the most overworked professionals out there because the work is complex and consequential. AI helps with speed, drafting, and first passes, but it doesn’t replace thinking. It just changes where time is spent. AI amplifies the strong ones. A lawyer with solid fundamentals uses it to think better, move faster, and focus on higher-value judgment. But if your foundations are weak, AI doesn’t fix that. It just helps you make mistakes at scale and exposes gaps sooner. Curious to hear from lawyers and operators here. Where have you seen AI genuinely help, and where does it clearly fall short?

Well said. Clients rely on professionals to carry risk, not just produce documents. AI can help, but it doesn’t take responsibility or exercise judgment. People still do that.

There is a consistent set of observations from people who actually understand BOTH law and technology, and this is one of them: “You don’t engage a lawyer to write a document. You engage a lawyer to hold the responsibility for your case. When something goes wrong, it’s not the output that gets questioned, it’s the judgment behind it. Accountability, context, and risk don’t disappear because drafting got faster.”

Colin S. Levy hits the nail on the head. If you're using AI purely to automate, draft, summarize, you're using it wrong. Using it as a collaborator and thinking partner is where you're going to get the most value.

The most effective way I have seen lawyers use generative AI is not as a drafting engine, but as a thinking partner with clearly defined limits. Collaboration works best when the lawyer controls how the task is framed and the AI handles an initial pass on structure, coverage, or recall. That might mean asking the system to surface issues to review in a contract, outline arguments before briefing, or summarize a record before deeper analysis. The value is not in accepting the output. It is in reacting to it. Collaboration starts to fail when the division of labor is unclear. When AI is asked to “draft a motion” or “research the law” without constraints, the lawyer is left reviewing blindly. A better approach is to use narrow prompts tied to discrete steps in the work, followed immediately by verification. Think sections rather than documents. Questions rather than conclusions. Citation checking illustrates this well. AI can accelerate research by quickly assembling cases and themes, but lawyers should assume they will validate every authority and refine every argument. What lawyers need are examples that mirror real practice: how to collaborate on discovery summaries, how to pressure-test an argument outline, how to use AI to improve clarity without giving up control. Abstract training about “what AI can do” does not help when you are actually at the keyboard. The teams that get the most value treat collaboration with AI as a skill. They decide in advance which tasks are appropriate, how outputs will be reviewed, and when the tool should be set aside. That clarity is what turns AI from a novelty into a reliable part of legal work. For those already collaborating with AI, which legal task has benefited most from this kind of step-by-step partnership, and which has been harder to make useful? I’m Colin, General Counsel of Malbek CLM for the Enterprise and author of The Legal Tech Ecosystem. #legaltech #law #learning #legaloperations #innovation

AI Looking into 2026: What Law Firm Leaders Should Be Paying Attention To As we turn into 2026, one thing is clear: AI has not hit a wall. Despite the noise around bubbles, regulation or “AI fatigue”, the reality inside law firms tells a different story. AI capability continues to advance rapidly and crucially, confidence at leadership level is materially higher than it was at the start of 2025. From what I see advising law firms, banks, regulators and professional-services organisations globally, three shifts matter most. 1. AI is no longer experimental - it’s structural We’ve crossed the threshold from pilots to embedded capability. AI is now: - Used weekly (often daily) by large parts of the workforce - Trusted for complex analytical tasks, not just summarisation - Quietly reshaping how junior and senior work is allocated - Even if AI progress froze tomorrow, organisations would still face years of operational disruption simply integrating what already exists. For law firms, this means AI should no longer be treated as "just" a LegalTech tool. It needs to become part of the firm’s approach to both the business and the practice of law. 2. The real shift in 2026: from prompts to agents The next phase is not better chatbots. It’s agentic AI: - AI that runs workflows, not just answers questions - AI that is assigned tasks, checks its own work and escalates issues - AI embedded into matter lifecycles, compliance processes and knowledge systems This is where most firms are not ready - legally, operationally or culturally. In law, this raises hard questions about accountability, professional responsibility, privilege, confidentiality and what “supervision” actually means in an AI-augmented practice. 3. Guardrails are no longer optional AI is already being used for legal reasoning, data management, and strategic decision-making. That makes AI governance a leadership obligation, not an IT problem. In 2026, the differentiator won’t be who uses AI. It will be who can use it safely, defensibly and at scale. That requires clear AI usage policies, training that goes beyond “how to prompt” and legal, ethical and regulatory alignment baked into design. Final thought AI is no longer a future shock. It is a present-tense management challenge. The organisations that win in 2026 won’t be the ones chasing the latest model. They’ll be the ones that rethink how work, risk and responsibility are structured in an increasingly AI- and data-driven world.

Most legal AI tools try to be one “general” agent that does everything. But real legal work doesn’t happen at the “general” level. An NDA doesn’t behave like an MSA. A limitation-of-liability clause doesn’t follow the same rules as a termination clause. Precision matters. That’s why CogniSync was built to support specialised AI agents - trained for each contract type, and even for specific clauses. And it gets even better: You don’t have to build them manually. Using automatic playbook and template extraction, CogniSync analyses your existing documents and distills the patterns inside them: → What your firm typically accepts → What it rejects → How you evaluate risks → How clauses interact with each other In minutes, the system turns your body of work into structured logic, ready to power custom agents that behave like your team, not like a generic chatbot. That means faster reviews, more consistent drafting, and AI that actually understands the difference between good, acceptable, and out-of-policy. Legal AI shouldn’t be one-size-fits-all. With CogniSync, you get AI that fits the way your firm thinks, negotiates, and decides. Test our platform → www.cogni-sync.com