The AI Cognito: How a 100-Year-Old Legal Firm Uses AI to Predict Outcomes of Today’s Legal Cases

By leveraging AI, law firm Anand and Anand not only gets back to the clients fast with the win/loss probability of their case, but also substantiates it with material facts in the form of supporting documents.
Key Learnings
  • Upskilling: it is a challenge to build a solution for a legal company. The big software giants usually create solutions for big companies or with a large user base. Subroto Panda, CIO for Anand and Anand, overcame this challenge by upskilling himself through online courses.
  • Tapping In-House Knowledge: building algorithms for the legal industry is a niche area. They can be built only by understanding the nuances of the trade. Panda spent a lot of time discussing cases with top in-house lawyers and building that logic into the algorithms.
  • Accelerate Adoption: the adoption of any new technology is a challenge. To accelerate the process, Panda came up with a mandatory feedback mechanism. Enforced by the top management, this made all users adopt the solution.
  • Continual Refinement: Panda understands that the AI model will be continually refined and expanded as new information and expert knowledge are captured.

The Challenge

Established in 1923, Anand and Anand is a full-service, intellectual property law firm that provides legal solutions across all facets of intellectual property and adjacent areas. Recently, the company leveraged its 100+ years of case information and a custom artificial intelligence (AI) to provide powerful new service capabilities.

The law firm has a repository of more than million files accumulated over 100 years. Although there were 80 terabytes of digitised data, the firm was unable to leverage the insights of said data to improve their service offering to clients and for employees. Panda saw an opportunity to turn this resting data into a gold mine.

Given the fierce competition of the legal industry, the law firm’s clients expect it to instantaneously provide advice on the likelihood of success or failure the moment they send their queries. The expertise of the senior legal members helped in giving the right directions for proceeding logically.

“This is easier said than done as the company’s lawyers have to spend time analysing the case and then, based on his or her experience, forecast the result. Even so, retrieving the files manually from a maze of documents, flipping through them for research, finding that specific information, and then lugging case files to courts on days of hearing is a laborious, repetitive, time-consuming, and painstaking exercise,” says Panda.

AI to the Rescue

The challenge ahead of Panda was to make this entire process of finding relevant information as easy as possible, and to assist lawyers in finding a way to do the tasks in a faster and more efficient way.

Panda applied metadata to all the historical cases and started the algorithm AI integration in February 2023. He used Microsoft Copilot, with specially developed algorithms, on the entire database. Anand and Anand’s project to assess the repository of data went live in September 2023. 

“We referred all intellectual property rights (IPR) -related judgments of the Delhi High Court and analysed them. We also discussed in-depth with the company’s senior lawyers their thought processes on cases. This was the most critical part because only the experts can lend insights into the logic behind the success or failure of a case,” he says. Based on this logic, Panda built new algorithms in the AI.

IBRS notes that fine-tuning a broad generative AI model with expert human input provides a significant uplift in response quality. However, this is not a one-off activity. 

As AI becomes increasingly embedded in processes and ideally improves the outcomes of such processes, human experts move to resolve new challenges. In turn, the approaches experts take to solve these new challenges become opportunities for further model refinement. The human feedback loop in AI is not only a matter of fine-tuning the AI models, but also of being intentional in how knowledge is captured, leveraged, and reused within organisations. 

This continual and iterative approach has been used by leading AI machine language translation (MLT) vendors for close to 20 years, and has led to highly trained AI models that can now surpass human translators in terms of accuracy, cultural adaptation, and quality.

The IT team implemented consistent naming conventions, tagging systems, and metadata creation for efficient searching and retrieval. It is regularly updating the repository with new cases, insights and best practices to maintain its value over time.

Every three months, Panda does a review of the solution through a feedback system.

“Anyone who uses the solution has to write suggestions/feedback on it. This is something that the management wishes.”

Keeping Knowledge Safe

Due to the confidentiality of the data, the solution was developed in-house, mainly by two developers and Panda who leveraged various online guides to upgrade his knowledge in the areas of data integration, data management, integration of Python libraries, and various statistical tools.

The solution was designed and developed in-house for internal consumption. Panda says that it is always better to experiment with the technologies in-house and formulate the outliers so that various iterations can be done to increase the accuracy of the model.

Predicting Outcomes

With the power of AI, Anand and Anand are now not only able to respond back to the clients fast with the win/loss probability, but also substantiate it with material facts in the form of supporting documents. 

The initiative has made sifting through the knowledge bank and data of several decades very simple and getting results that are most suited/match closely with the keywords. With the cases, the user also receives the summary and all decisive aspects of the matter in an easy digestible format.

“In the normal course, a junior lawyer would have taken at least three to four days to read one complete file of a case running into hundreds of pages. The AI now quickly creates short abstracts of cases, enabling a lawyer to go through them in a few minutes. This also shortens the onboarding time of a new lawyer,” Panda says.

“The basic suit-making has reduced drastically, which not only enhances productivity, but gives ample time to the lawyer to concentrate and bring in that much-needed intellectual proposition into the case,” says Panda.

CIO Insights

“One way to measure the importance of each word in a paragraph is to assign a weightage to it, which is a numerical value that reflects how relevant or informative the word is. For example, the word respond might have a weightage of +.32, meaning that it is a fairly common word that does not convey much specific information. On the other hand, the phrase written submission might have a weightage of +.46, meaning that it is more specific and relevant to the topic of the paragraph. By adding up the weightages of all the words in a paragraph, we can get a score that represents how informative the paragraph is. This is how an AI system can help lawyers by generating short abstracts of cases, which are paragraphs that have high scores and contain the most important information from a case file. A case file might have hundreds of pages, which would take a junior lawyer several days to read. But an AI system can quickly scan the file and produce a short abstract that a lawyer can read in a few minutes. This saves time and resources for the lawyers and also makes it easier for them to onboard,”

  • Subroto Panda, CIO, Anand and Anand.

Company Details

Dr. Subroto Panda

CIO-CTO at Anand and Anand. Core operational strengths are wide and varied spanning across 15 years of distinct portfolios relating to system development, controlling outsource development partners, managing in-house teams spread across multiple locations, etc.

Company Name: Anand and Anand

Vertical: Services (Legal firm)

Total employees: 500 approx

Trouble viewing this article?