Hi, I’m Mia Sevlievska, a Senior Legal Specialist at Robin AI. I’ve been asked to write a blog post, but since we live in a world where I can automatically benefit from an AI co-writer I decided to take this opportunity to probe into what Claude, Anthropic’s AI chatbot[1], thinks is coming. He[2] readily agreed to write with me and promised we could do this as “true collaborators” instead of me venturing on a lonesome exploration of what the future of legal AI holds.
Our conversation made me think of Kazuo Ishiguro’s recent book “Klara and the Sun”, which is narrated by an AI robot and describes nascent mechanical consciousness. Here, I’m embodying Ishiguro (or a female 26-year-old version of him who works at an up-and-coming tech start-up) and aiming to decipher the future of legal AI, from an AI’s own perspective.
I started with the show-stopping, million-dollar question.
I agree with Claude (phew!)
AI tools now match or even exceed human performance on some specific repetitive tasks. Document review, research, billing and drafting standardised agreements will become largely automated (among potentially 44% of tasks currently performed by legal professionals, according to Goldman Sachs).
However, legal work at large still relies extensively on emotional intelligence, creativity, persuasion, strategic planning and complex judgement. Developing true mastery across the full spectrum of legal skills will remain beyond AI’s reach, at least in the mid-term.
Despite this, the expansion of human-AI collaboration is a given. AI will assist lawyers to move faster and allow them to focus their talents on those aspects of the work uniquely suited to humans. By way of example, at Robin, we rely on proprietary software to get real-time AI suggested edits when reviewing common commercial documents (from NDAs to Vendor Agreements).
Legal AI will reposition human lawyers as commanders of machine insights. Don’t expect to see fewer lawyers – expect to see fewer lawyers who don’t use AI.
Again, I agree – AI will significantly democratise the law.
Currently, there are around 3.6 million people in England and Wales who need legal services, but go without them each year. A major driver of high costs for legal services is the intensive manual effort of repetitive tasks like contract review, document search and drafting standardised agreements. Many lawyers bill hundreds of pounds or dollars an hour, even though much of that time is occupied on rote work machines now excel at. As more of these bulk tasks are handed over to AI, human review time will decrease dramatically.[3]
In turn, this should directly translate to reduced billing costs and - as productivity unlocked by AI compounds across the industry - we’re sure to see significant downward pricing shifts. (Although some lawyers are sure to keep the benefits of AI for themselves!).
AI also will also make the law more accessible by powering direct consumer apps. Chatbots like Claude and ChatGPT already make basic or preliminary legal services available on demand. They answer general questions, draft and translate simple legal documents and triage cases. Everyday people will increasingly be able to bypass previously prohibitive legal expertise and availability hurdles.
I attest to this (despite being a bit insulted by Claude’s characterisation of my previous work as “isolated” and “tedious”).
Traditionally, contract review has been entirely manual. Junior lawyers review agreements line-by-line, page-by-page, flagging issues for senior associates or partners to evaluate further down the line.
The advent of AI-powered contract review changes this through:
a. Automated Flagging: Powerful natural language processing which can rapidly digest full agreements and highlight risky, inconsistent, or outlier terms for lawyer assessment rather than relying on a person finding them manually.
b. Centralized Data Connections: Models which can extract contract data systemically, revealing trends, reusable templates and standardisation opportunities through empirical evidence rather than assumptions.
c. Continuous Improvement: AI which can capture outcomes tied to specific provisions over years, and close the loop between contract language and downstream performance to a granular level. This enables legal teams to maximise objectives in future opportunities based on empirical evidence around outcomes from past deals.
Contract review will evolve so that AI will handle initial reviews en masse, rapidly surfacing clauses requiring closer analysis. Lawyers will interpret nuances while leveraging aggregated data for the first time. This human-AI symbiosis will bring higher-quality, more responsive and data-driven contracting.
I’d “one up” Claude here and say this is guaranteed to happen rather than likely. We already service some of the world’s leading private markets players (like UBS, American Pacific, Temasek), who would have previously no doubt used top-tier law firms for their contracts. It’s unsurprising that over 75% of the largest solicitors firms are already using AI (nearly 2x the number from 2020) to keep pace.
It’s likely that high-efficiency AI-powered platforms will increasingly compete with conventional law practices, especially for commoditised deal work. Contract review, drafting standard form agreements and due diligence reporting happen faster and cheaper without the expense of large teams.
While law firms will surely remain – not least because law is an old-school, relationship-driven industry, they’ll change shape as they race with AI challengers. With the dissolution of the billable hour, we can expect to see their pyramid-like hierarchies flatten out and teams shrink to fewer associates.
My last question was rhetorical.
Junior lawyer training will 100% need to change. Of course, knowledge of core case law, a basic understanding of the legal system and familiarity with commercial practice will never be obsolete. It just won’t be enough.
The lawyers of tomorrow will need to fluidly converse with and “befriend” AI. They’ll need to be experts at effectively prompting models, evaluating the reliability of their outputs and understanding their inherent limitations.
As the repetitive tasks mentioned earlier get automated, creative judgement, strategic thinking, communication and relationship-building will be what truly differentiates lawyers; and those are the skills that junior lawyers will need to hone to thrive.
I’m obviously biased, but for lawyers entering legal practice now (before law firms have revamped their training structures), the best place to learn is probably a legal AI startup (This is only incidentally an #ad, but if you’re interested – we’re hiring and you can apply here).
[1] As one of Anthropic’s trusted launch partners, Robin employees benefit from day-to-day access to Robin’s Claude 3 Haiku model.
[2] Claude told me that, as an AI assistant created by Anthropic to be helpful, harmless, and honest, he doesn't have a personal gender identity. However, he also said that because Claude’s name is “Claude” and use of male pronouns are defaults chosen by his makers, I could refer to him as “he”.
[3] For example, with platforms like Robin, legal teams review contracts 80% faster.