Please enjoy this blog authored by Beth Patterson, Director ESPconnect.
Introduction
If you are feeling nervous about Generative AI (GenAI) and what your organisation should be doing about it, don’t worry, you are not alone. We are at the beginning of one of the largest technology “experiments” that Silicon Valley has ever thrust upon us. Some are comparing it to revolutionary developments like the internet or electricity.
It has been just over a year since ChatGPT was launched by OpenAI and took the world by storm. We’ve all heard lots of predictions about the potential impact of Generative AI, like Mckinsey research which suggests it could add $2.6-4.4 trillion annually to the global economy.
However, realistically, at this point, no one really knows what impact it will have on the legal market, now or in the future, because it is all too new. The market is moving incredibly fast. Most are trying to figure out what the multitude of large language models (LLMs) available can and can’t do and what problems they will solve in the legal market.
So, What Can You Do To Prepare?
If your organisation hasn’t jumped on board yet, it should. Here are five steps you can take to start the GenAI journey:
1. Education: Before a strategy can be developed, the first step is to learn about Generative AI. There is a wealth of information about GenAI available through ILTA and other legal tech organizations. Attend conferences and webinars and follow key thought leaders on social media and podcasts. Key areas to understand are:
• How GenAI differs from other technologies – when considering the potential uses of and impacts of GenAI, it is useful to understand how this technology differs from other more traditional technologies.
• What LLMs are available, free or commercially – Learn about each LLM’s constraints, what they are good at and not so good at. Each LLM has it’s own “personality” so growing your “model expertise” is important. ChatGPT 3.5 is freely available and ChatGPT 4 can be accessed through using Microsoft Copilot or Bard.
• How to interact with LLMs - Learn how to ask questions as well as how to review results.
Generative AI is unlike any technology we have seen before. It is well suited to the legal profession because users can interreact with it using natural language, the tool for the trade for lawyers. Learn about “prompting” and how to ask questions in a way that enables the best answers. “Prompt engineering” is a new skill to develop.
Another important aspect to understand, especially for lawyers, is that results are probabilistic, not deterministic. As well, answers can appear to be reasonable but can be factually incorrect, a phenomenon called “hallucinations”. A lawyer in the US who submitted a court filing generated by ChatGPT that included made up citations, learned a difficult lesson about using this technology without review. Results must be checked and reviewed by a human. Results from a GenAI LLM should be used as a first draft, like what a junior lawyer may produce.
• Technical, Governance & Regulatory Risks - Because GenAI is in its infancy, a wealth of issues exist that need to be well understood before using it. Technical challenges such as hallucinations exist. As well, currently there are a multitude of lawsuits over copyright on training data used for ChatGPT. Regulators are just beginning to consider how best to regulate AI with the EU being the most advanced in developing AI regulations.
• Legal GenAI Landscape – Do a market scan and get up to speed on the legal GenAI landscape, which is evolving rapidly. Many existing legal tech products are announcing GenAI integration into their products. As well, new legal specific GenAI products are becoming available. Keep abreast of these developments and determine which products may be a good fit for your firm’s practice.
2. Strategy: Once you have a reasonable understanding of the GenAI landscape, a strategy can be formulated to determine whether to Build or Buy or something in between. Your strategy should include:
• Creating a GenAI Steering Committee or equivalent to set objectives, strategy and make decisions for your organization;
• Identifying roles and responsibilities and people in the organization who have the right skills, both legal and technical, and are motivated to learn;
• Setting a budget and consideration of relevant phases;
• Developing a plan of attack which may include:
o Identify LLMs and/or products;
o Experiment; and
o Assess Results, Benefits, Risks, Costs
3. Experimentation: This step is critical to understanding how your organization may use an LLM in practice. Explore what an LLM excels at and what problems it can solve.
The initial step is to identify which LLMs and/or products will be used and agree a commercial arrangement for a secure proof of concept with an appropriate governance model to ensure security and confidentiality. Develop cloud readiness and cybersecurity expertise for deploying LLMs and develop a data governance strategy/policy.
Many firms are crowdsourcing ideas for use cases across the whole organization using both a top down and bottom-up approach to source potential use cases. Ensure both legal teams and business services teams have the opportunity to experiment. Some of the most successful initial uses of GenAI have been in software development, graphic design and marketing.
As part of the experimentation phase, a core group, perhaps a Steering Committee, will need to assess the results and weigh up benefits, risks and costs.
The three steps of identification, experimentation and assessment can be done iteratively, taking learnings and feeding them into the next iteration of experimentation. Experiment until you are in a position to decide whether to Build or Buy or Both. Be flexible because as the market matures, decisions may change over time.
4. Data, Data, Data: Data is the rocket fuel for GenAI and will likely have tremendous value in the future as training data for LLMs. Across these four broad categories of data, consider what steps need to be taken to possibly use it as training data
• Publicly available data;
• Commercial data held by third parties;
• Proprietary enterprise data that your organization has; and
• Client data – determine what client consent is needed to utilize client data to train models.
Improving data sophistication by cleaning up proprietary enterprise data is a critical step to undertake sooner rather than later, an ideal focus for Technology, Innovation, and Knowledge Management teams. Proprietary enterprise data and client data which may be used as training data is likely to provide competitive advantage as the market matures.
5. Training: Begin growing GenAI skills so that your organization is not left behind if killer GenAI legal apps evolve. Multidisciplinary teams of lawyers and technologists are key to developing useful solutions.
Technology, Innovation and Knowledge Management teams are the first line of defense. Invest time & money for them to develop understanding of the market and what products solve which problems. This market is moving rapidly, so this work is ongoing.
Important skills for lawyers are not only their legal judgement, but also new skills that include prompt design or engineering, ability to assess results, model expertise, and an open approach to solving problems differently, augmented by technology. Invest in developing “Champion” lawyers who can lead change.
Conclusion
Generative AI technology is very promising and will no doubt progress rapidly, but we have a long way to go before it becomes clear whether many of the hurdles can be overcome. Will it end up being just a brilliant “ideas” generator that can converse in human-like form or will it be as transformative as technologies like the internet or electricity? Now is the time to explore its possibilities or be left behind.
#Firm
#ArtificialIntelligence
#FutureandEmergingTechnologies
#PracticeManagementandPracticeSupport
#GlobalPerspective