Anand & Anand is a leading 100 plus year, full-service law firm in India, widely regarded for intellectual property (IP), corporate, advisory, and dispute resolution work. The firm holds particularly deep expertise in trademarks, patents, designs, and copyright law, serving a broad range of clients ranging from small businesses to some of India and the world’s largest enterprises.
As part of leveraging cutting edge-technology, the company decided to leverage AI. However, AI was not brought into the law firm as a futuristic trial. Rather, it was implemented out of necessity, to create a solution that addressed the critical challenge of effectively meeting customer expectations with the best possible response time (especially during off-hours and at sensitive moments).
The first major use case was trademark filing support. Anand & Anand’s business involves the entire spectrum of intellectual property – patents, designs, trade marks, copyright. A lot of work involved at the stage of filing is rule based, repetitive and time sensitive.
Previously, for example, when a client from the US sent in a last-minute mail with a request towards the close of Friday in India, response work would start only Monday morning. This delay proved to be a bottleneck for business. Dr Subroto Kumar Panda, CIO at Anand & Anand, decided to maximize AI to overcome this challenge.
Building AI from Inside the Firm
Anand & Anand did not outsource the core AI capability. The solution was built in-house by a very small team of two people, both of whom have been with the firm for 15 years. Panda says this was a deliberate decision. “If we bring in external people, there are chances that even if we sign the NDA, you cannot delete a human memory,” he says.
The project took about six months to develop and went live in April. To support the workload, the firm had to upgrade its infrastructure and procure new Nutanix servers with higher GPU capacity. Panda says that was essential because “LLMs and the RAG model require a huge GPU.”
Today, as the system processes the mail, it picks up the relevant data, matches it with the filings standards of the trade marks office, and generates a draft reply that the legal team can look up, to complete.
The firm has already deployed the solution to more than 100 people. The response, Panda says, has been positive because employees could immediately see the benefit. “They were appreciative because it gave them more time to do other things, to do a lot of research, and to concentrate on the more technical aspect with respect to any case,” he says.
From Delayed Responses to Near Real-Time Service
Panda says the impact was immediate once the firm started using AI and agentic workflows in client servicing. “Everybody expects an instantaneous reply,” he says. “With AI and agentic AI, the email is being read, the required information is formatted in the particular format, and it is uploaded. The client receives an email.”
That turnaround, Panda explains, would earlier have taken nearly two days if the request came in on a Friday and the office was closed for the weekend. Panda says the change has had a direct impact on service quality. “The waiting time has come down drastically,” he says. “What used to take 14 hours or even two days now takes only a couple of hours.”
The system is also designed to work within the firm’s operating model. Since legal services require caution and accuracy, nothing is sent out automatically. Instead, the AI prepares the draft, and a human reviews it before anything goes to the client or the filing system.
Why Human Oversight Still Matters
For Anand & Anand, the biggest principle behind the AI rollout is human control. Panda is clear that legal work cannot be handed over entirely to a model. “Human in the loop is the most important principle in our AI policy,” he says. “Nothing should go out without human intervention.”
That approach matters even more in a legal practice, where information is highly confidential and errors can have serious consequences. The firm therefore chose to build applications that can be containerized inside its own environment rather than relying on external LLM platforms. “We are more concerned with developing applications which can be containerized inside our permissions,” Panda says.
The firm also invested in internal guardrails to ensure that sensitive information does not escape its controlled perimeter. Panda explains that the team can visually see how data is processed, how it is vectorized, and how it is matched against templates and nearest vectors. This transparency helps the team understand exactly how the system is working and where boundaries need to be placed.
Lessons From This Model
The biggest takeaway of Anand & Anand’s AI experience is that to get the right fit in a machine learning model, you need a solid process framework that predates and dictates. Before building a model (even a simple one), the team needed to scope the problem, triage the information (or data), clean it and test prototypes to get it right before scaling up. Only then could it set up and define templates to help automate the extraction and ensure that the entire process remained compliant.
Another key lesson is the importance of choosing use cases with visible business value. In this case, the payoff was measurable: a filing process that once took three to four hours now takes about 10 minutes. “Just imagine,” Panda says. “On a Friday, the processing would have started on Monday. Now the total process time has come down to just a couple of minutes.”
The broader impact goes beyond speed. Panda believes the firm has improved client experience, reduced employee fatigue, and freed up time for more meaningful work. He estimates that the business impact could eventually translate into more than 5 percent improvement, driven by faster service and higher client satisfaction.
Panda says that AI should not be treated as a trend to satisfy boardroom pressure. It should be used to solve repetitive, fatigue-inducing work, but only with strong guardrails. “The people developing AI applications should be completely immersed in the project,” he says. “They should have the desire to see how their information is being processed.”
For legal and knowledge-led firms, Anand & Anand’s example shows that AI’s real value lies not in replacing expertise, but in amplifying it safely, transparently, and at speed.
