Maveric Systems Wants Banks to Walk Away From Flashy AI Pilots

The banking, financial services, and insurance (BFSI) industry is still consumed with GenAI deployments; however, they rarely translate into real outcomes. As per an EY study, 74% of Indian financial services firms are running AI pilots, but only 11% of them have reached production.  

Maveric Systems is trying something new. It has chosen to step back from the churn and redraw the terms of engagement.

The Chennai-based firm, a specialised BFSI-focused technology company working with global banks, regional leaders, and select fintechs, is consciously resisting what Co-founder and CEO Ranga Reddy describes as the industry’s default AI playbook: the indiscriminate pursuit of proofs of concept.

“We don’t go and explore any use case and every use case that comes from the customer,” Reddy tells AIM. “We have predetermined and said that in five areas we have the ability to assist you.”

Those five areas define the boundaries of Maveric’s AI strategy—applying AI to data quality and data validation, radically transforming software quality assurance, transaction monitoring, accelerating software development life cycle, and building what it calls a “knowledge fabric” under its AI-powered document intelligence solution, Maveric DSprawl.

The 26-year-old company positions this as a deliberate rejection of the industry’s tendency to chase fragmented use cases.

Maveric’s AI Strategy

The company has two major streams—core services and AI services. 

While the ₹550 crore ($57 million) core services segment is growing at 6–7% annually, revenue in the AI segment has grown from about $1 million in the earlier phase, when much of the work involved unpaid proofs of concept, to about $3 million last year. The company is targeting around $9 million this year and approximately $17 million (around ₹150 crore) in the following year. 

Much of it comes from areas such as data validation and software quality assurance, where traditional services are transitioning to AI-enabled delivery models, as well as from platform-led engagements.

Maveric’s decision to focus on execution rather than fragmented pilots is grounded in what it has observed over the past two years. In the last one year period, the company ran 25 proofs of value across these five areas. Only six moved into production. “It was easy to do experiments, but not very easy to actually produce anything at scale,” Reddy admits. 

The bottleneck is not technological capability but the way banks structure and govern AI initiatives. That’s why Maveric has chosen to tighten its execution model. It demonstrates its platform in a generic context, without embedding enterprise-specific knowledge.

“Give me 12–14 weeks, and I can actually show you very specific proof of value,” Reddy claims. The intention is to ensure that each deployment is grounded in measurable outcomes rather than exploratory breadth.

The company is prioritising investments to scale its platform-led AI business, but not through the public route, which Reddy believes remains “very, very lukewarm”. Instead, it is looking to dilute its stake by around 26% to raise capital. The company expects to bring a global systems integrator on its cap table before the end of the financial year.

Knowledge Fabric

At the centre of this approach is the concept of the knowledge fabric, which Maveric sees as critical to making AI usable in banking environments. Reddy draws a distinction between data and knowledge, arguing that conventional data preparation is insufficient. 

“Knowledge does not reside in one format,” he says, pointing to the need to bring together structured and unstructured data, internal and external sources, documents, policies, and regulatory inputs. While knowledge graphs form a component, the broader objective is to create a unified layer that contextualises AI.

“AI comes with public information… putting a layer of enterprise knowledge on top of it so that anything that AI throws out is more contextual as compared to generic,” he explains. In practice, this means building domain-specific knowledge fabrics, for instance, in lending. That contextual layer is then consumed by downstream AI systems.

This shift from generic to contextual AI is now driving demand. Among the banks that have progressed beyond experimentation, Maveric is seeing multiple parallel initiatives centred on knowledge fabric creation, whether for regulatory reporting, internal reporting, lending, or trade finance. “That is the most popular and… much in demand,” he notes.

Drawing a Line

Maveric also has a layered partnership model across technology and consulting firms. 

On the one hand, it partners with large technology providers such as Microsoft and Google, which its banking clients typically use for cloud infrastructure, AI infrastructure, and tooling. In addition, the company works with specialised AI product firms aligned to specific lines of business, such as trade finance or lending, where “AI products… create an agentic approach to running the trade finance business or lending business,” Reddy says. 

Beyond technology partnerships, Maveric also collaborates with banking-focused consulting firms, which help define business parameters and transformation roadmaps. These firms assess what outcomes banks want to achieve from Maveric and outline changes across process, people, and technology.

Once a proof of value demonstrates results, clients can either adopt the platform through a licensing, implementation and support model, or move to a transaction-based pricing approach, particularly across four of the five solution areas. The knowledge fabric itself is treated differently, as it does not directly generate outcomes but serves as the foundational layer for subsequent AI-driven processes.

This disciplined approach has also led Maveric to take a harder stance on client readiness. The company now refuses to proceed with AI engagements unless three conditions are met within the bank: an enterprise-wide AI governance framework, a clearly prescribed and accessible set of AI tools and infrastructure, and defined business parameters that AI initiatives are expected to impact. 

These checks are embedded in the sales process. “The first three questions they will ask you are: do you actually have an entire governance process established… have you already prescribed the set of AI tools and infra… and does the line of business clearly give the direction and the scope with regard to what business parameters AI initiative should impact,” the CEO says.

Why AI Adoption is Shaky

The stance reflects a broader diagnosis of why AI adoption in banking has struggled to scale. 

Reddy points to three recurring gaps observed over the past 24 months: the absence of unified governance, a tendency to focus on low-risk internal processes rather than customer-facing applications, and a lack of clarity on the business outcomes AI is meant to drive. Banks end up pursuing pilots that deliver limited impact and fail to clear compliance thresholds when they approach production.

There are, however, signs of a shift. Around 35% of banks, by Maveric’s estimate, have begun to put these foundational elements in place, including standardised tool stacks. 

For Maveric, the implication is clear. By refusing to engage in unstructured experimentation and insisting on readiness before execution, the company is betting that discipline—rather than volume—will define the next phase of adoption.