chevron-down Created with Sketch Beta.

Law Practice Today

March 2024

The Year You Hire a Chatbot

Jared Jaskot

Summary

  • Lawyers can use chatbots to interact with clients in innovative ways.
  • Chatbots need to be carefully deployed, fine-tuned, and tested to avoid potentially serious errors.
  • Law firms must understand the data privacy policies of any chatbot vendors they hire.
The Year You Hire a Chatbot
iStock.com/diego_cervo

Jump to:

2023 was the year in which chatbots and AI became the buzziest topic in legal technology and for good reason. A new class of artificial intelligence called large language models demonstrated ability with words and ideas across a broad range of domains that appeared to surpass humans in many ways. A chatbot powered by GPT4 passed the bar exam in March 2023. For the first time, many people began to believe that AI would actually change the face of the law. Chatbots offer many skills to attorneys, including summary, drafting translation, research, analysis, transcription, and more. This article will focus on utilizing chatbots in your law firm to interact with your clients and give some simple rules to get the most out of these new tools while avoiding embarrassing and costly mistakes.

I have been building and utilizing chatbots to interact with clients in my law firm for eight years. Chatbots have many advantages over human staff for certain tasks. Chatbots work 24/7 and can have thousands of conversations simultaneously. They speak multiple languages and do not tire of repeating the same conversation over and over again. Chatbots are a truly scalable workforce that can respond to staffing needs instantly. A successful marketing campaign or a viral video on social media can often lead to a large volume of contacts. Scaling up instantly to handle increased volume is nearly impossible for human staff. Software and cloud computing easily handle big swings in contact volume efficiently and economically.

A growing number of customers prefer chat to phone, and as younger generations become consumers of legal services, this number will expand. Chat has a number of advantages for law firms. It gives customers the feeling of instant responsiveness. It can handle FAQs, appointment settings, reminders, and some parts of intake. Chatbots dramatically outperform most websites in conversion rate from advertising. The ability to improve marketing, sales, intake, and customer service with a relatively low-cost solution should cause all law firms to investigate this solution for their own use.

In 2023, attorneys jumped on the AI train with enthusiasm, and there were mixed results. An attorney in a Federal Court in New York was sanctioned for a brief he submitted using the tool after it came out that his brief was full of citations to cases that do not exist. In other domains, a hacker convinced a car dealer’s chatbot to sell him a car for $1. 2023 showed that, despite their tremendous potential, chatbots can have major downsides when not deployed thoughtfully. 

Human employees occasionally make mistakes. Law firms have learned over time how to train and deploy staff in spite of inevitable mistakes. Chatbots should be approached with a similar mindset. The elimination of chatbot mistakes is a goal that can be achieved through a fine-tuning process geared toward guard rails and intelligent conversational design. Primary goals for law firm marketing chatbots should be to provide common information during the intake process, to triage prospective clients, and to help clients schedule appointments. Problems that need to be avoided include off-topic or sensitive conversations, legal conclusions, and offering extensive legal services and/or the prices of those services. It can be helpful to think of front-line intake staff and the responsibilities and authority they have when delegating authority to a chatbot.

Conversational design for chatbots is an important consideration. I wrote an article in 2020 about ten chatbot design principles that I still believe are relevant today. The first rule, “Don’t be greedy,” seems to be the most difficult for lawyers to understand when they create chatbots. First, seek to understand your customers’ needs and goals in the conversation before you begin to push them about information you want. There is no rush when it comes to chatbot conversations. Customers often talk to bots two or three times before they make a buying decision. Understanding the customer journey is critical when we design and evaluate chatbot performance.

Off-topic conversations need to be trained out of your legal chatbot. It would be extremely embarrassing for a law firm to receive screenshots of an inappropriate conversation with the firm’s chatbot. Large language models can and will have extremely inappropriate conversations. A well-trained bot can avoid this problem by refusing to discuss any topics outside the legal domain in which the law firm works. This can be a challenge if the firm handles matters with sensitive subjects. More extensive training and testing should be used in these circumstances.

Chatbots need to avoid making legal conclusions that will mislead customers and expose law firms to liability. Bots must be trained to always use phrases like “it appears that” or “it is likely.” Bots should identify themselves as such at the outset of a conversation and also indicate that they are not lawyers and cannot make legal conclusions. Bots can avoid customer expectations of a legal conclusion by avoiding extremely detailed back-and-forths that are likely to lead to a legal conclusion.

Chatbots should not be empowered to make offers to clients for legal services outside of scheduling consultations. In my firm, bots have had the ability to charge for consultations but have rarely been able to make the sale without some human intervention. If firms want chatbots to talk about the price of legal services, it should only be in the context of a range of prices. Bots should never commit a firm to representation.

Chatbots training must ensure compliance with the code of attorney ethics. Bots should not guarantee success. Bots must avoid language that compares the attorney’s performance to other attorneys or that implies the attorney is a specialist (unless allowed). Many sales and marketing bots have the capacity to engage in outbound marketing campaigns and extensive follow-up sequences. Attorney ethics rules may prohibit these activities, and attorneys must ensure that the bots they hire curtail these capabilities while they work for the law firm.

Law firms need to understand the data policies of the tools they use to deploy chatbots. Major vulnerabilities in customer information must be understood and avoided. Law firms need to understand what third-party vendor employees have access to customer conversations and data and what steps prevent them from accessing said data inappropriately. Most third-party AI vendors’ technology relies on external services that they license. Law firms must understand how their data is protected when it is passed to these external services. The only surefire way to protect customer data is to be thorough in investigating and understanding these complex data agreements.  

The process of fine-tuning and training your law firm chatbot to the above goals needs to be done thoughtfully. There are many vendors who are capable of this task; make sure that you vet them and ask about their methods and experience with sensitive data and conversations. Off-the-shelf chatbots are rarely up to par for the law firm without a fine tune. Saving money by skipping this step and using a premade bot is a mistake. A poorly trained bot will miss business opportunities, give your firm a bad look, and potentially expose you to ethical issues or liability.

Test your law firm chatbot! The only way to know that your bot complies with all of your goals and restrictions is to test it and attempt to induce errors. Law firms should commit to a thorough testing regime before deployment and should regularly engage in ongoing testing as AI models are quickly evolving. Force your software provider to provide you with a testing plan and show you the results of their fine-tuning and testing. Testing a bot before and after fine-tuning will help you better understand how bots can train to become better employees.

Talk to your law firm chatbot at least once a week. Do you like the conversation? How do you want the bot to represent you and your firm? Is the bot achieving this goal? This technology is highly malleable. Keep working on your firm’s bot and it will improve over time and help you achieve your goals.

    Author