Ending Tax Notices: Clear Believes AI-Native Data Lakes Can Shield Enterprises

The finance teams of enterprises in India seem to always be a step behind when it comes to tax compliance. Some of it can be attributed to just the sheer volume of transactions, especially when it comes to managing indirect taxes such as input tax credit and implementing e-invoicing. 

However, with changing regulations and legal rulings in areas such as transfer pricing, categorising business expenditure, treaty benefits, and the treatment of non-compete fees, enterprises find themselves chasing a flurry of tax notices. According to a Moneycontrol analysis, there has been a 1,000% surge in tax defaults over the last decade, with unpaid direct tax demands for corporate entities estimated at ₹3.97 lakh crore in FY23.

Large companies file “roughly 200 tax returns per year”, with government systems increasingly cross-verifying filings and issuing automated notices. “It takes roughly a week for finance teams… to find the data needed to respond to the notice. That is one notice… they get 200 of these notices,” Swaroop Repaka, Senior Director of Product at Clear, tells AIM, adding that notices can arrive “up to four years after you have reported to the government”.

Fintech company Clear (formerly ClearTax) believes the solution cannot be AI adoption alone, but ownership of the underlying data layer. It recently launched Direct Tax AI capabilities within its Clear Compliance Cloud platform, as part of a broader strategy to build a unified foundation for enterprise tax compliance. The company said it has expanded its total addressable market by 70% with the move.

“AI is as good as the context, the data,” Archit Gupta, Founder of Clear, said during the launch, underscoring the company’s focus on integrating data across tax workflows before layering AI on top.

This unified approach, he said, addresses a core enterprise pain point: fragmentation. “They’re not looking at different data points from different point products… the fragmentation of the data platform presents a lot of compliance risks, which we’re taking away by having one unified offering,” Gupta noted.

Clear’s Data-First AI Strategy

Repaka says the company’s AI strategy is built on a data-first architecture. “Without a foundation such as a data lake, you cannot build, you cannot automate anything,” he notes.

At the core of the platform is a finance data lake that aggregates information from enterprise systems, government portals, and internal workflows. On top of this layer, the company is deploying a combination of classical machine learning and LLM-based systems. Clear’s transaction intelligence engine uses machine learning models tailored to each enterprise. 

“For every customer, we have to train that additional model for them… on your data, we will train the model,” Repaka says, noting that differences in tax structures and policies make standardisation difficult. Clear is adding new AI-led use cases, including tax notices, litigation, refunds, and subsidies.

These models learn historical transaction patterns to flag anomalies such as incorrect tax treatments, inconsistent vendor classifications, or deviations in TDS rates. “Once they have a model, it will then detect if that transaction goes out of whack,” he says.

However, AI-led tax compliance systems also invite regulatory scrutiny.

“AI solutions, being probabilistic in nature, bring the critical challenge of managing hallucinations. While difficult, this is essential to minimise errors without converting the system into a purely rules-based one,” Debashish Banerjee, Partner, Deloitte India, observes.

Strong governance ensures the AI does not fabricate past cases, generate incorrect new sections, and that the context behind any response remains fully traceable. “Equally important is data governance, including compliance with geographical regulations and the implementation of robust technological safeguards," he adds.

Also, designing the system with expert architects and implementing two to three layers of risk governance ensures that AI-driven compliance decisions remain defensible during audits. “While outputs from LLMs or SLMs may not allow every word to be traced, establishing the source of context—through review bots or expert oversight—is crucial,” Banerjee asserts.

Centralised Data

In parallel, the company is building agentic AI workflows for finance teams. “We are building a similar platform for the CFO office,” Repaka notes.

The impact of this architecture is most visible in handling tax notices—one of the most resource-intensive aspects of enterprise compliance. In FY24 alone, GST officials issued 1.13 lakh show cause notices due to data discrepancies.

With a centralised data layer, response times can drop sharply. “If all of this data is on a data lake, you can respond to any notice in a matter of hours,” he says, adding that “AI also makes it easy for you to reply to that notice in a very automated manner.”

Clear’s push into enterprise AI-led compliance builds on regulatory digitisation trends following the introduction of GST in 2018. The company began with indirect tax compliance and later expanded into direct taxes, with that product line initiated in 2021.

Clear Compliance Cloud for Indirect Tax offers capabilities such as automated data preparation from ERP systems, end-to-end reconciliation across financial and tax records, and AI-powered matching of government and enterprise data to improve accuracy and unlock working capital. The platform also helps enterprises proactively manage compliance risks through vendor scoring, real-time discrepancy detection (including e-way bill mismatches), and automated handling of notices and filings.

Repaka argues that this segment has seen limited innovation from both enterprise software vendors and services firms. “It would have been natural for someone like SAP or Microsoft to address this… but it has largely been ignored,” he observes.

With its data-first platform now in place, Clear is also taking the model global. “We are already in Malaysia… we are going to Europe… we have many logos,” Repaka said.

The company said it has closed two to three large enterprise deals in the latest quarter and expects more in the pipeline, as it looks to position its compliance data stack as the foundation for enterprise AI in finance.