Legal Tech Trends Newsletter: #14
UI for AI, Intake Forms, GenAI in Kira, Fundraising and much more!
Happy Friday, and welcome to the 14th edition of Legal Tech Trends!👋
I’m back from a hiking trip in Italy and excited to add three twists to the newsletter:
A mini-series on how law firms are securely accessing large language models.
Interview insights from leading law firm tech and innovation teams.
A media partnership with the Legal Innovators conference.
It’s fun exploring topics and newsletter formats that resonate with me. When time permits, you can expect the return of the short audio segment in future editions.
This edition is a bit meatier than previous newsletters. If you like diving deeper into topics, you’ll enjoy it. But please do let me know what topics and newsletter formats you find most interesting.
Let’s jump in!
UI* for AI - Secure AI access for law firms
Part 1 of 2
(*UI = User Interface)
We’ve seen a flurry of law firms announce customised ChatGPT-based offerings in recent weeks. Addleshaw Goddard, Dentons, Gunderson Dettmer, Allens.
Way back in the 5th edition of this newsletter, I shared a simple graphic to help structure our market understanding. I’ve updated it below to show where these recent announcements fit and highlight some law firms using this option.
It’s exciting to see firms taking concrete and considered steps to use the latest AI securely, but as the AI hype wave continues, it’s easy to lose sight of what these announcements mean and the current state of play.
This topic is important, so to help bring clarity to the discussion, I spoke to Shawn Curran, Director of Legal Technology at Traver Smith, and Michael Kennedy, Senior Manager (Innovation & Legal Technology) at Addleshaw Goddard.
These were fascinating chats, and I could easily fill this whole newsletter with their learned insights; however, I’m conscious of information overload, so I’m sharing the snippets most relevant to this topic now and will release the full consolidated takeaways as a separate blog post in a future edition.
⚠️ First, let’s distil the recent announcements ⚠️
Law firms are not building their own large language models or attempting to recreate ChatGPT.
They are not fine-tuning or refining these foundational models.
Most law firms are accessing the same, unrefined, models. Including GPT-3.5 and GPT-4 in Azure OpenAI Service.
These models have the same AI capabilities and limitations as the models you can access directly in ChatGPT.
So what’s happening? How is this different to ChatGPT and what’s already available?
ChatGPT is an incredible consumer product to easily access OpenAI’s models via a chat interface, but enterprise software has drastically different needs from consumer software. If I use ChatGPT to summarise an article or brainstorm birthday presents, I don’t really care what happens to the data or if the output is slightly wrong. However, large law firms care deeply about this.
These firms handle sensitive client data and have a valuable brand reputation to protect, so they must enforce stringent security and privacy requirements.
As mentioned, I could easily fill this newsletter multiple times over, but here are the top three takeaways I think you’ll care about most, which are related to the recent law firm LLM-related announcements.
Shawn Curran
🎨 Three takeaways related to creating custom UIs for LLMs.
Travers Smith generously launched their open source YCNBot several months ago. This is a fantastic free resource to help other firms rapidly deploy their own UI to access large language models.
They have released additional features such as:
Case law detection: Can’t copy and paste case law.
Removal of PII: Detect and block personal information.
Prompt sharing: Good prompts can be promoted to the "Community".
The primary purpose of YCNBot is to educate and experiment. They are still taking a very cautious approach, and deeply assessing the risks.
Michael Kennedy
🎨 Three takeaways related to creating custom UIs for LLMs.
One of the primary aims for the Addleshaw Goddard product is also to test, learn and educate teams about LLMs. They are exploring several products.
A chat interface is great for certain use cases, such as querying and surfacing functionality of products like Word or Excel which users may not be aware of, but in many use cases a chat interface may be less suitable. They are exploring default prompts and presets which are actioned via buttons.
The chat interface may be an ideal location to surface additional functionality to users and interact with other non-LLM related tech available.
It’s hard to discuss large language models in a law firm context without highlighting the risks. Although many have been well covered, these additional risks and scenarios raised by Shawn are worth sharing with you now.
🌟 Risks to make you think twice…
OpenAI Copyright Lawsuits: OpenAI has several copyright related lawsuits against it from content owners. OpenAI has clearly scraped websites which had legal contracts available, without the permission of those sites. If a law firm generates contract content from an OpenAI model, which will be based on legal content they did not have permission to access, and if OpenAI loses the copyright lawsuits, how may that impact the validity of the contracts which used OpenAI models? Could law firms that generated content with OpenAI for contracts see those contracts void? Maybe the odds of this are very low, but it's a scenario which could play out.
DMS Data Ownership: Law firms store contracts from many sources in their DMS, including contracts drafted on their own paper, other law firms' paper, and clients' paper. Which of these contracts do they have the right to use if they want to train their own LLMs or fine-tune 3rd party LLMs? Will they have to filter the dataset to ensure they own the data, and prove that they have not used data they do not own in their training data?
GPT-5: When GPT-5 is released, it will likely have been trained on the data users input to ChatGPT. When users first started using ChatGPT, they did not realise the data input may be used to train future models. Many people likely input sensitive data, which could now end up in GPT-5. This could be a major issue when it is released!
End of Part 1.
The second part, on the 1st of September, will feature insights from two additional leading law firms that are exploring how to safely access large language models.
Round-up
📢 LexisNexis announce more details of Microsoft collaboration
Lexis Connect, their legal intake and matter management product, which works within Microsoft Teams and uses conversational AI will launch in Q3 of this year. They will also have a plugin for Microsoft 365 Copilot. Given Copilot will cost $30/user/month it’ll be curious to see the charging model for various plugins. LINK
🌟Litera unveils new gen AI capabilities in Kira, starting with summaries
They’re starting with summaries using GPT 3.5, available from August 26th. The ability to find information in contracts using natural language questions and requests will be added in the coming year. Given Kira’s expertise in AI contract review, their approach is particularly interesting. They still believe their existing Kira tech is better than Gen AI in 80% of use cases. LINK
💸 Harneys sells fiduciary business to private equity giant Hillhouse
The offshore firm will reinvest a “significant portion” of the sale proceeds into the law firm by creating an accelerator fund to invest in new technologies and products, explore new markets and optimise business operations. LINK
📃Online Intake Forms: lessons learned
Emerging tech hogs the headlines, but there are so many other obvious improvement opportunities in legal. Online intake forms are one such area. Alex Hamilton collated his learnings in one useful document. LINK
🧱 Betty Blocks AI - The low-code way to secure and flexible AI solutions
Betty Blocks has launched a feature set to allow customers to easily access and explore various LLM models in their low code environment. They are especially focused on enabling less technical users to swap the LLMs they access easily and launch products faster. LINK
Fundraising & Deals
💸 Syncly raise seed funding led by Stuart Barr, former chief product & strategy officer of HighQ
Syncly provides real-time synchronization between legal platforms such as iManage, NetDocuments, HighQ, and Microsoft Teams. LINK
The HighQ angle is particularly interesting to me as:
The vast majority of the Top-50 UK firms use HighQ.
Syncly is seeking to solve known HighQ DMS integration limitations.
They want to proactively monitor and shut down HighQ sites that no longer need to be active.
💡Adjacent idea - Storage requirements influence HighQ costs so compressing files before uploading can reduce costs. If you’re exploring an automated solution for this please let me know.
💰 BriefCatch Raises A $3.5M Seed Round to Expand Its Legal Editing Software and Add AI-Powered Features LINK
💵 The Contract Network raises $8m from several high-profile investors
The AI-powered contract collaboration platform parses out contracts into their individual clauses, and uses generative AI, including GPT-4, to provide context and market insights for each change and redline. LINK
Podcasts
🎙️ EE 277 - The Barrister Turned Businessman Disrupting An Entire Industry - Dan Fox
The founder of Johnson Hana, a leading Irish Alternative Legal Service Provider, discusses his entrepreneurial rollercoaster journey. Most folks agree that being a lawyer is a demanding career, so it’s particularly nice to see the flexibility his alternative model offers. LINK
🎙️Pioneers and Pathfinders - Wendy Butler Curtis
Wendy Butler Curtis is the chief innovation officer and chair of the eDiscovery & Information Governance group at Orrick, Herrington & Sutcliffe LLP. This podcast provides great insights into the innovative approaches Orrick is taking and the influence of other non-legal related disciplines. LINK
Adjacent Interests
💡World Economic Forum - Top 10 Emerging Technologies of 2023 LINK
⚠️ Open challenges in LLM research LINK
💫 McKinsey unveils its own generative AI tool for employees: Lilli LINK
Media Partnership - Legal Innovators
The Legal Innovators conference occurs in London on the 8th and 9th of November. The conference will bring together the industry's leading law firms, in-house legal teams, legal services businesses, and legal operations professionals.
As part of this media partnership with Legal Tech Trends, you can expect to find interesting content related to the conference in upcoming newsletter editions. This may include insights from conference speaker interviews and other content that resonates.
Aaand that’s a wrap. The next edition drops on Friday, 1st September. In the meantime, to satisfy your curious legal tech mind, explore LegalTechTrends.com.
Thanks for reading. All feedback and suggestions are most welcome.
Have a great Friday!
Feel free to connect on LinkedIn for additional interesting content Here 👋