Tally Solutions rewrites core software for AI

That is economically significant because accounting software sits close to the operating heart of small and mid-sized businesses. If AI can reliably ingest invoices, bank statements and other documents, it can reduce manual entry, speed reconciliation and cut back-office labor. That lowers the cost of compliance for customers, but it also changes the economics of software vendors that have historically sold stability and trust rather than experimentation.

Tally has already started moving. In May, it launched TallyIra, an AI-powered product that uses Google’s Gemini and Chinese open-source model Qwen with OCR to extract data from invoices and bank statements and convert them into accounting vouchers. Nabendu Das, chief of engineering and products, said the tool can process documents in different formats and reduce the need for human data entry. About 20,000 to 25,000 customers have used the product since launch, covering roughly 6 million to 7 million pages of invoices and statements.

But the early adoption figures also underline how unfinished the transition remains. TallyIra is still a separately purchased add-on rather than a native feature inside TallyPrime, and users need a licensed version of Release 7.1 or later plus an active Tally Software Service subscription. That suggests Tally is still testing demand, pricing power and product fit before it makes AI a default part of its core platform.

The company’s leadership is candid about the stakes. Chief executive and founder Bizstone? [No] — managing director Phani Goenka described the redesign as “painful” but necessary, saying the firm has repeatedly junked old code rather than keep patching it. He said Tally began AI pilots about 18 months ago, but added that the technology remains at a nascent stage and that “someone is going to figure this out” across incumbents and startups alike.

For investors, the key question is whether AI can extend Tally’s franchise or whether it will force a costly rebuild with uncertain payoff. The bull case is that the company’s long customer relationships, installed base and compliance credibility give it a strong position to monetize automation. The bear case is that AI features could be commoditized quickly, while the cost and complexity of rewriting a mature accounting stack could slow execution and create product risk.

There is also a broader labor implication. If invoice capture, voucher creation and bank reconciliation become more automated, demand may shift away from routine bookkeeping toward review, exception handling and higher-value advisory work. Goenka acknowledged the disruption, though he said he does not know how it will affect individual workers. That matters in a market such as India, where millions of small businesses still rely on Tally as the backbone of their accounting workflows.

Accuracy remains the central constraint. Because Tally’s software is used for tax filings and sensitive financial records, a probabilistic AI system cannot be allowed to act unchecked. Das said the company is keeping a human in the loop so AI can flag possible issues rather than automatically create vouchers. That is the right safeguard for customers and regulators, but it also means the near-term productivity gains may be incremental rather than transformative.

Tally says its ownership structure has allowed it to keep hiring even as AI spending rises, and Goenka said the company’s AI bill is “fairly large.” That points to a familiar pattern in enterprise software: the near-term margin story can worsen before the long-term product story improves. Whether Tally can navigate that trade-off without losing trust will determine if its AI-first rewrite becomes a growth engine or an expensive act of self-disruption.