Since 2017, Diab Chambers LLP, a specialty law firm with over 30 years of experience focusing on mass disasters, has recovered over $1.65 billion in settlements on behalf of its wildfire clients. The firm has represented over 100 public entities – including some of the largest cities and counties in California – as well as thousands of individuals and families in their cases against utility companies for damages or injuries caused by wildfires. Most recently, the firm represented the County of Maui in the devastating Maui wildfires. Founding Partner Ed Diab, who actively uses Personal AI, was awarded the Daily Journal California Lawyer – Attorneys of the Year (CLAY Award) in 2020 for some of these historic results in wildfire cases.
Why did Diab Chambers choose Personal AI?
Executive testimony from Ed Diab
Perfect recall and accuracy
“The beauty of the product is the high level of accuracy – again, because it is trained on MY data and not a large language model.”
Preservation of institutional knowledge
“Personal AI allows us to profitably pass on the knowledge base of a busy Senior Partner or a retiring senior employee to the cohorts below.”
“LLMs are extremely powerful tools that effectively attempt to index the universe of knowledge on the internet. Personal AI, instead, indexes your universe of knowledge. Trained solely on your memos, your briefs, your writing style, and your firm’s approach, its recall is as good as the data provided. Thus, this ‘personal language model’ (“PLM”) provides an extremely high degree of accuracy and confidence because it is effectively pulling from your outsourced memory.”
Privacy and security compliance with the rules of professional conduct
“From a privacy and security standpoint, Personal AI has features that differentiate itself from the publicly available LLMs.”
Challenge #1
Accuracy and Accessibility of Information
When a new client comes into a firm, there invariably are dozens of other previous cases handled by the firm that included similar facts, issues of law, discovery disputes, deposition transcripts, or expert witnesses. The talented litigator or law firm is one that finds an area(s) of law that it becomes highly specialized in and does it repeatedly, with great success.
1 In Mata v. Avianca Inc., a U.S. District Court for the Southern District of New York judge sanctioned two New York lawyers in June 2023 for submitting a legal brief generated by ChatGPT. The brief contained citations to six fictitious cases.
The status quo requires the litigator or staff on the case to do several things:
Think about what cases were similar – either the set of facts or issues of law;
Have the staff “pull” those cases (i.e., search the firm’s database or case management system in a manual fashion);
Have the staff “pull” deposition transcripts of experts that have testified on similar cases;
Search email inboxes; and
Search the case management system for relevant documents.
This list is by no means exhaustive, but the status quo essentially involves a search across fragmented platforms that usually relies heavily upon the recall of the lawyer or staff member. And even once the relevant document, transcript, or court order is located, it then requires a lawyer or staff member to read through the document to find the desired information.
Solution #1
Personal AI eliminates time spent on a variety of these workflow and communication barriers
The platform sits across the entirety of the lawyer’s, staff’s, and/or firm’s systems. Boundary conditions can be easily created to define what data sets it should be trained on. On one side of the spectrum, an “all in” approach can be taken – training on emails, SMS messages, Slack or Teams threads, motions, pleadings, court orders, etc. On the other side, a more surgical approach can be taken, albeit still automated, to train on curated data sets.
In practice, this allows the lawyer to interact in a natural language fashion with the entirety of that lawyer’s digital footprint – whether in written form or recorded audio or video. And perhaps most importantly, the accuracy and reliability are extremely high (potentially, perfect) because it draws on that lawyer’s discrete database of information. The potential for “hallucinations” is heavily mitigated.
Challenge #2
Preservation of Senior Employee Knowledge
Law firms universally struggle with the challenge of passing down knowledge and information to new members to the firm – whether lateral hires or young Associates. Frankly, this is a problem that all businesses with more than a handful of employees face, which is how to profitably spend time passing on the knowledge base of a retiring, senior employee to the cohorts below. Lawyers don’t get paid to train other lawyers. Contingency firms are incentivized to focus their efforts on prosecution and settlement of cases. Billable firms are incentivized to work on behalf of their clients, and therefore bill, and produce revenue for the firm. In neither instance are lawyers incentivized to take time out of their day for the “non-billable” task of training. To be fair, larger law firms have made tremendous progress in this regard, but it still takes a tremendous amount of time, energy and resources to do properly. For most small to medium sized law firms, however, training is usually done on an ad hoc basis, if at all. “Trial by fire” continues to be a mantra all too widely used amongst law firms. A useful vignette is this all-too-familiar situation at a firm. A second-year Associate is working on a motion on behalf of a twenty-year Partner . The Associate asks the Paralegal on the matter to pull all prior motions that were written on similar cases.
This includes the opposing party’s briefing as well as the ultimate judge’s orders in the cases. Perhaps there were some deposition transcripts of relevant testimony that were utilized to form the factual bases of the motions. The staff then needs to pull those transcripts for the Associate to review as well. The Associate diligently reviews all of the aforementioned documents, conservatively spending six to seven hours reviewing them. The Partner on the case who will inevitably review the Associate’s bills may need to cut those hours because “our clients don’t pay for you to learn on the case,” the Associate is told. Seven hours is cut to two billable hours. The time the staff spent to pull the relevant documents is perhaps cut as well, if it was even captured to begin with. For any practitioner, the above-referenced situation is a daily occurrence. And six to seven hours is extremely conservative; i.e., it likely will take that Associate a couple days of work to review the material. After the Associate’s review is complete, the Partner is met with a familiar knock on their door, or perhaps in the post-COVID era, a Zoom invitation. The Associate has questions. More non-billable time is taken.
Solution #2
Instead of taking up time from other staff members, Associates can have their questions answered immediately.
Enter, Personal AI. The Associate asks, “I have [this type of case]. How have you all handled a motion to strike [an applicable cause of action] in previous matters?” A synthesized response – not only from the singular twenty year Partner , but all Senior Partners’ databases at the firm – is generated within seconds. The Associate then proceeds to ask, “Which depositions were relied upon in the [ABC] matter” generated in the previous response. And so on, and so forth. The Associate communicates in a natural language fashion with not only one Partner , but all Partner s, or perhaps three of the Partner s that the Associate feels have the best subject matter expertise. The Associate can ask the lawyers’ digital twins which documents were relied upon to more accurately and quickly search the firm’s database of institutional knowledge.
Cases the twenty-year Partner had long forgotten from seventeen years ago are suddenly recalled. As the saying goes, “the senior Partners have forgotten more than the Associates know.” The foregoing represents a limited use case of Personal AI’s potential. Realistic potential use cases include a generative component – i.e., utilizing Personal AI to generate written content, responses to emails, portions of briefs, deposition outlines, etc. The platform can also be used to preserve a departing lawyer’s knowledge base, recall what was discussed in a Zoom meeting with a client, and essentially any other situation where “perfect recall” would be useful.
Challenge #3
Privacy and Security Compliance Using AI
Comment 8 to the American Bar Association’s (ABA) Model Rules of Professional Conduct, rule 1.1 provides: “To maintain the requisite knowledge and skill, a lawyer should keep abreast of changes in the law and its practice, including the benefits and risks Associated with relevant technology, engage in continuing study and education and comply with all continuing legal education requirements to which the lawyer is subject.”
Moreover, states’ rules of professional conduct are rife with lawyers and law firm’s obligations to maintain the privacy and security of their client’s confidential data. They are required to make reasonable efforts to prevent the inadvertent or unauthorized disclosure of information relating to the representation of a client. This often expressly includes taking appropriate steps to protect electronic data and communications. The rules further provide that a lawyer must utilize secure methods of communication and storage and should stay up to date on cybersecurity best practices.
Solution #3
Personal AI has two features that differentiate itself from the publicly available LLMs.
First, Personal AI’s Terms of Service expressly provide that the customer owns and controls the data. To be precise, “Your personal AI works by training on your private digital memory vault, including photos, text and other uploaded materials, referred to as a Memory Stack. This allows your AI to learn more about you and better predict your needs and preferences. The larger your Memory Stack is, the more information your personal AI can remember about you (collectively, the “Personal.ai Services” or “Personal AI”). Memory Stack data is owned by the user and will not be sold to third parties or used for interest-based advertising.”
The Personal Language Models are trained solely on the customer’s data, ensuring it stays private and secure. Second, each Personal Language Model is separate and unique to the user, providing an additional layer of security. This means, among other things, that Personal AI does not use any one customer’s data to help train its models for the purposes of another customer. A hypothetical breach would only impact that specific model unlike LLMs where a single incident could expose vast amounts of data.
“The beauty of the product is the high level of accuracy – again, because it is trained on MY data and not a large language model. “
Ed Diab Founder, Diab Chambers LLP
Summary of Key Results
Personal AI offers a unique and secure way of creating personalized language models that can help lawyers and law firms improve their productivity and quality of work. Unlike Large Language Models that rely on massive amounts of data from various sources, Personal AI only uses the user's own data and preferences to create a tailored and private model. Moreover, Personal AI adheres to the highest standards of security and ethics, as detailed in the Big Law Firm Ethics Whitepaper.
.svg)
