How Zentis AI makes BFSI automation actually work
With specialised AI models, a no-code studio, and early traction across global banks and insurers, Trivandrum-based Zentis AI aims to become a top AI automation player for the BFSI sector.
Across banks and insurance companies, digital transformation often promises efficiency but delivers frustration. Amidst AI pilots stalling and automation breaking down midway, teams fall back on manual work. Regulations shift faster than systems can adapt, and data is rarely clean enough for smooth deployment.
For Deviprasad Thrivikraman, Managing Director of Zentis AI, these problems weren’t theoretical; he had lived them for over three decades working across global BFSI operations.
“More than 90% of the AI projects in BFSI fail not because the technology is weak, but because they’re not battle-tested for the industry,” Thrivikraman says. Zentis AI, incorporated in April 2025 in Trivandrum, was born from this view of what truly blocks automation in the BFSI sector.
Over the last six months, the startup—incubated at Techvantage.ai—has been working on what Thrivikraman describes as a “BFSI-native agentic automation platform”. It uses lightweight, CPU-friendly specialised language models to automate core processes in banks and insurers without demanding heavy GPU infrastructure.
“We realised no one was solving the real problem, making AI safe, compliant, and truly adoptable,” Thrivikraman says.
Thrivikraman’s 30-year career spans companies like Cognizant, Satyam, and Wipro, with stints in the US, the Middle East, Europe, and the UK. Before Zentis AI, he spent 10 years as CEO of Techvantage.ai and had already built and exited other ventures, including a peer-to-peer insuretech platform with the Allianz accelerator.
The startup’s main challenges were choosing the right problems to solve, finding a team that understood both BFSI and AI, and convincing big banks to trust a new product.
Some clients had weak tech systems or messy data, which slowed deployments. Working across many countries also made compliance and processes harder to manage.
What Zentis AI does
At the core of Zentis AI is its no-code Studio—a simple drag-and-drop canvas—where analysts can design automation workflows without writing code.
Once a workflow is created, it can run in two ways: Zen Pilot, which acts like an assistant for processes handled directly by people, such as internal audits; and Zen, a silent agent that works in the background for continuous tasks like claims or loan applications.
The system is powered by lightweight, BFSI-focused specialised language models that run on basic CPU machines, making deployment easier and far more cost-effective for enterprises that cannot invest in heavy GPU setups.
The platform is also LLM-agnostic and cloud-agnostic, allowing banks to stay within their Microsoft, AWS, or GCP environments.
For many clients, Zentis deploys through Docker on on-premise or air-gapped private cloud setups, since regulated industries rarely move sensitive data into SaaS platforms.
To speed up the adoption, Zentis offers a growing library of ready-made BFSI workflows—from underwriting to claims processing and internal audit—so clients don’t need to build from scratch.
Working with banks and insurers
According to Thrivikraman, deployment takes less than a week once system access is cleared. The platform ingests documents, checks for missing information, flags mismatches, and drafts reports or decisions for human teams to validate.
Today, Zentis AI works with 25 active clients; 10 have already moved into paid engagements through a POC-to-paid model that uses clear, agreed-upon performance metrics. This provides the startup with an impressive 83% conversion rate.
Without divulging names, Thrivikraman says the startup’s clients include three long-established banks in the UAE, two of the world’s top five insurance carriers, three mid-sized Indian banks, and a major insurance broker in the UAE, including RAK Bank.
The bootstrapped startup, with a provisional valuation of $10 million, operates with an eight-member team, expected to grow to 10 soon. It expects to raise funding soon, with formal fundraising discussions beginning early next year.
Revenue model: licensing and outcome-based pricing
Zentis uses a straightforward licensing model, charging $3,000–$4,000 a month for each automated process. This can sometimes go up to $5,000 if the workflow is complex, coming to nearly $40,000 per year per process.
Thrivikraman reveals that some clients in the UK are also trying an outcome-based model, where they pay per claim processed, which he sees as a strong future option.
For SMEs and brokers using the SaaS version, pricing averages around $1,500 per month, although uptake is slow due to data-sharing concerns. To address this, Zentis plans to launch a secure private data room in early 2026.
Zentis positions itself differently by focusing on a BFSI-native agentic platform, something still missing in most global solutions. It competes with horizontal automation platforms like E42 and language-model players like Sarvam AI. While they offer strong tools, they are not built specifically for banking and insurance.
What’s next?
Zentis is now preparing for its UK debut at the Insurance Innovators Conference, where it will showcase its platform to a wider audience for the first time. It is also planning to enter the US market through a channel partner and has already signed a reseller agreement to expand into Saudi Arabia.
It aims to strengthen its global presence and work towards its target of becoming one of the top 10 AI-driven automation players in the world within the next five years.
According to the Precedence Research report, the market size of global digital transformation in the BFSI sector is $108.5 billion in 2025, projected to reach $419.4 billion by 2034. “We’re aiming for around 10% of the global AI automation market in BFSI within five years,” Thrivikraman tells YourStory.