Legal design is AI readiness
For eight years, we’ve been redesigning contracts, terms and policies so people can actually understand them. It turns out we were doing something else at the same time: getting those documents ready for AI. Here’s why the two disciplines were always one.
The market finally has the words
What we mean by the “information layer”
Since 2018, Inkling has practised legal design: restructuring legal documents around the people who use them. Plain language. Logical structure. Tested comprehension. When we told the market this was about the information layer (the quality and structure of the knowledge itself) most buyers heard “make the documents prettier.” But Inkling had been working with AI already. Inkling was an IBM partner and had built a number of AI solutions for itself. Inkling knew the powerful benefits well structured documents would have and the critical issues lawyers would face when trying to combine AI with complex legal documents.
Fast forward to 2026 and organisations are trying to put Copilot and AI assistants on top of their document libraries, and have now discovered what the information layer actually is: it’s everything. The AI’s answers inherited every ambiguity, every buried condition, every 70-word sentence in the source. Suddenly the qualities legal design has always produced — clarity, structure, tested understanding — became the difference between an AI that gives correct answers and one that confidently launders confusion at scale.
Legal design has always been about useability of law. AI readiness is it’s prime use case.
Why legal design and AI readiness are the same thing
What makes a document work for AI is what makes it work for people
An AI-ready document and a human-ready document are built from the same materials:
Clear structure: AI retrieval works on chunks — sections that make sense in isolation. So does human navigation. A document with logical headings, one idea per clause, and consistent hierarchy is easier for a person to find their way around and easier for an AI to retrieve accurately. Legal design has been building this structure into contracts for a decade; retrieval engineers now call it chunking strategy.
Plain language: Ambiguity doesn’t disappear when an AI reads it — it gets summarised, fluently, with unearned confidence. A clause a person can misread is a clause an AI can misrepresent. Plain language drafting, the core craft of legal design, reduces the error rate for both readers at once.
Explicit conditions: Buried cross-references and implied exceptions are where human readers get lost — and where AI answers silently drop the qualifier that mattered. Surfacing conditions, another legal design staple, is retrieval accuracy by a different name.
Tested comprehension: Legal design’s discipline of user testing — does a real person, doing a real task, reach the right understanding? — is precisely the evaluation AI deployments are missing. We now run the same benchmarked testing on the whole pipeline: source document, AI answer, human decision.
The market treats “AI-ready content” and “plain language” as different budgets owned by different teams. They are one property of one asset: your knowledge, designed for its readers. Both of them.
Inkling’s approach to AI Readiness
We didn’t pivot to AI. Inkling was founded on it and it’s been part of our DNA from day 1.
Inkling’s founder, Sara Rayment, started Inkling Legal Design after seeing the power of AI as a delegate at the United Nations (UNCITRAL) in 2016. She returned home, partnered with IBM and started a small AI consultancy called Inkling. Hoping to sell AI solutions, she quickly discovered that most firms weren’t AI ready; their content was too complex, they had no clear workflows and their governance was a mess. Inkling began focusing on the information layer to get these firms ready for the extensive automation that AI would bring. Inkling focused on the AI LLMs, the lawyers creating the systems and the end users to ensure the information was accurate from end to end.
Since that time, Inkling has completed over 150 projects. The Financial Times has independently assessed our work since 2021 — and read in order, the record shows one discipline maturing, not a firm chasing a trend: recognition first for legal design itself (FT Innovative Lawyers winner, 2021), then for psychographic techniques that test how readers actually understand documents (2023), then an FT award for redesigning ANSTO’s commercial agreements around non-lawyer users (2024), then for our AI solution built on that testing data (2025), and most recently for our AI-powered predictive drafting tool, which puts years of human user-testing insight to work inside an AI system (2026). Human understanding first; AI trained on the evidence of it. That’s the order the work has to happen in. See our full recognition record →
What this means for you
If you’re rolling out AI over your documents, we can help.
The uncomfortable truth in every Copilot deployment: the tool is rarely the problem. The knowledge is. Before you tune prompts or switch platforms, ask three questions of your document estate:
can the AI retrieve the right content (structure);
will its answer be understandable and correct (clarity); and
does anyone verify that real users reach right decisions (testing)?
That’s an information-layer audit, and it’s where every engagement of ours begins — whether the client arrived searching for “legal design” or “AI readiness.” Same discipline. Same fix. Two readers, one outcome.
FAQ section
What is legal design?
Legal design applies human-centred design to legal documents, services and systems — restructuring contracts, terms and policies around the people who use them, and testing with real users that they’re genuinely understood.
What does “AI-ready” mean for legal documents?
An AI-ready document is structured and written so AI systems can retrieve the right content and represent it accurately: clear sections, plain language, explicit conditions and consistent terminology. The same qualities that make it human-ready.
Why does AI give wrong answers from our policies and contracts?
Usually because of the source, not the tool. AI answers inherit the ambiguity, buried conditions and complexity of the documents they read — then deliver them fluently and confidently. Fixing the knowledge fixes most of the answers.
How do you test whether documents work with AI?
We test the whole pipeline with real people: the redesigned source, the AI system reading it, and the human acting on the answer — measured against realistic tasks and benchmarked against our global user-testing portfolio.