Earlier this year, a lawyer in the US, overwhelmed by work pressure, turned to ChatGPT for assistance, but the chatbot generated false scenarios, resulting in the lawyer’s dismissal.
While large language models (LLMs) can produce inaccurate information, they are still undeniably a valuable tool for legal professionals.
Jhana.ai, a Bengaluru-based startup, recently launched an AI-powered legal research and drafting tool, Jhana, for the Indian practice. The co-founders claim they’re building an “intelligent legal system”.
AI Agent for Legal
Founded by Benjamin Hoffner-Brodsky, Em McGlone & Hemanth Bharatha Chakravarthy, Jhana recently raised $1.6 million from Together Fund with participation from Shyamal Anadkat from OpenAI, and Kunal Shah from Cred, among others.
“What we’re trying to build is a sort of paradigm shift that centralises all the datasets lawyers utilise, enabling them to be accessed and employed in various ways through AI,” Hemanth Chakravarthy, a co-founder at Jhana.ai, told AIM.
The idea is to help the lawyers at each step, from onboarding a new case and reviewing prior legal history, to preparing petitions and arguments for the court, and more.
“We have this paralegal, which is a human-like agent. You send it instructions of whatever nature – a loose query, a structured set of instructions, anything at all – it has a thought. It has many experts it can access on this, almost on a phone-a-friend basis,” Chakravarthy said.
Currently, lawyers rely on search engines and various journals and academic resources to locate case law and statutes. They draft and review documents manually, relying solely on their individual expertise.
Jhana consolidates all relevant datasets, enabling lawyers to interact with them seamlessly using AI. It also comes with a proprietary dataset, which is around 15 million documents and over one-and-a-half trillion tokens.
LLMs Powering Jhana
“Our philosophy contrasts with the approach of companies creating massive models from extensive datasets, claiming they can solve all problems – which often proves ineffective,” Benjamin Hoffner-Brodsky, a co-founder at Jhana.ai, told AIM. “Instead, we believe in a biomimicry approach, observing how advocates work.”
He further explained that their approach involves conducting research, collaborating with colleagues and asking questions instead of simply taking information and repeating it. He said the goal is to incorporate the most effective models for particular tasks, developing a workflow that reflects human methods.
The founders believe this approach helps keep the model grounded and prevents hallucinations. Additionally, every result generated by Jhana is supported by citations, allowing lawyers to easily identify the sources.
The tool utilises a range of models, including a mix of proprietary and fine-tuned open-source ones, along with paid API provider models. Their approach involves employing numerous specialised models tailored for specific tasks, such as extracting a table of dates from legal documents.
“That’s almost a part of our trade secret. We have expert models for every little task, like pulling out a table of dates from a legal document. And that’s part of how we tether our AI to truth rather than hallucination,” Chakravarthy said.
The startup has built its own version of Falcon, the open-source model developed in the UAE. Besides, it also leverages OpenAI’s GPT models. “We have also experimented with Antrophic’s Claude since they released Sonnet. However, historically, we found OpenAI’s models to be more suited for our use cases,” he added.
(Jhana AI co-founders Benjamin Hoffner-Brodsky, Em McGlone & Hemanth Bharatha Chakravarthy)
Who Can Leverage Jhana?
Brodsky mentioned that Jhana is designed so that almost all stakeholders in the legal community can benefit from the platform.
“We offer products tailored specifically for enterprise users, catering to both large law firms and in-house legal departments. These solutions include integrations with your private, siloed data, and on-premise options.
“For smaller firms and individual users, we have offerings designed to meet their needs. Whether you’re an independent advocate, part of a small firm, or a law student, our affordable individual subscription plans allow access to our extensive external data,” he revealed.
Currently, Jhana is available at INR 3,300 per month for the basic plan and INR 5,000 per month for the Pro version. For enterprises, the pricing remains similar but can be customised based on additional services, such as enterprise-tier data integrations. Interestingly, the company has decided to keep Jhana free for judges and judicial officers.
By August, the startup had onboarded over 500 users through invitation. This includes CNGSN & Associates LLP, which is one of the biggest tier-two law tax firms in Chennai.
“Another example is Reliance, as we are part of a JioGenNext Programme. We are currently collaborating on a procurement system and exploring ways to scale our efforts there,” Chakravarthy revealed.
Meanwhile, several other startups are developing similar products.
Earlier this year, Vinod Khosla-backed Sarvam AI launched Sarvam AI Legal, a generative AI workbench aimed at empowering lawyers. It offers features like regulatory chat, document drafting, redaction, and data extraction to enhance their capabilities.
Chakravarthy, however, was unimpressed. He expected better from a startup with over $40 million in the bank.
“I’ve seen products that perform a Google search using a LangChain API, which can be easily copied from a GitHub repository. They combine this with PDF OCR to enable interaction with PDFs, along with web searches and legal prompts, filtering results to only include legal sites. While I prefer not to discuss competitors, I believe that in fields like health tech or legal tech, such practices can be irresponsible,” he said.
Founding a Startup in India
Chakravarthy met Brodsky and McGlone during his undergraduate programme at Harvard. Brodsky moved to India in 2018 for an internship programme with J-PAL South Asia. He discovered that India was a place where he could make significant strides more quickly, encountering larger and more complex problems to tackle.
“I mean, in the US, where I come from, all the software-as-a-service (SaaS) startups are solving some tiny problem for some very small demographic. And the problems we faced here were big, foundational. They warranted the effort.”
When AIM asked the Americans about their experience in founding a startup in India, they had lots to say. Brodsky declared that Bengaluru was one of the best cities he had ever lived in. “I have to say, for the kind of things that we’re doing, which is starting high-growth tech companies, I don’t think I’ve ever been to a better city in the world.”
However, they did encounter a few challenges in the wake. For instance, Brodsky commented, “India is a country that loves stamping!”
His remarks highlighted the challenges of doing business in India compared to the ease of the process in the US. “We have a dual entity structure with both a US and an Indian entity for investment purposes. In the US, when you incorporate a startup, you simply pay a fee, click a few buttons, and within 24 hours, you have a legal corporate entity registered with the government—often without any direct human interaction.
“In contrast, the process in India requires a more hands-on involvement. You’ll need to hire professionals to assist you, which can take anywhere from one to four months. This process often involves in-person visits to your home or office, extensive paperwork with numerous stamps, and even the necessity of displaying a signboard outside your office,” he said.
On a personal level, Brodsky also revealed that the need for paperwork has made it challenging to even open a bank account in India. “ I actually gave up on trying to get, like, a driver’s licence and a bank account. It was just way too much paperwork. My co-founder Em has been trying for six months now,” he said.
As a result of not having a local bank account, he also shared his disappointment in not being able to leverage UPI. While there are offerings such as Cheq UPI – designed for foreigners to reap the benefits of UPI – one can use it to transact with a business but not an individual.
“You lose access to a lot of the beautiful micro-economies of India that involve paying someone INR 30 via your phone at a tea stall, which is kind of a shame,” he said.
The post Jhana AI is Set to Disrupt India’s Legaltech with AI-Powered Research appeared first on AIM.
