The next phase of artificial intelligence in financial services may be determined less by how much time AI saves than by what it makes financially and technically possible. Once increasingly capable AI systems begin addressing problems that were previously too complex, expensive or dependent on specialised expertise, the competitive advantage could shift from automation to the ability to attempt entirely new things.
That proposition emerged during the Global Fintech Fest 2026 session ‘AI Native Markets: Driving the Next Era of Financial Advantage’. The 30-minute session on September 10 formed part of a GFF 2026 programme examining the movement from technological potential to impact, alongside Agentic AI, tokenisation, quantum technologies and trusted, connected financial systems.
For Anthropic India Managing Director Irina Ghose, who spoke in the session, productivity remains important but increasingly capable AI could make efficiency alone a less distinctive advantage. If banks, insurers, asset managers and fintech companies can all deploy frontier models for documentation, customer support, reconciliation, coding and analysis, the more enduring advantage may belong to institutions that use the capacity created by efficiency to pursue new products, markets and problems competitors have not yet been able to solve.
She describes this additional capacity as “headroom” for innovation, leading to the larger “moonshots” she has urged business leaders to identify. As frontier AI becomes more capable while deployment costs decline, the economic proposition begins moving from automating routine work towards making difficult, specialised and previously impractical problems economically viable.
Financial services provide an especially important testing ground because institutions operate some of the most complicated technology environments in business. Applications, databases, software and operational systems accumulated over decades remain intertwined with critical processes, while institutional knowledge can be concentrated among a relatively small number of experienced employees.
Ghose's example of legacy banking code illustrates the potential transformation. Work that could previously require months and highly experienced personnel could potentially be compressed into days, while AI could help employees understand old architectures, identify dependencies and navigate unfamiliar code. The significance therefore extends beyond faster coding to the wider availability of specialised knowledge, although human validation, governance and accountability remain essential.
This is where AI Native Markets becomes more than a conference theme. There is a fundamental difference between placing an AI assistant alongside an existing workflow and redesigning the workflow around systems capable of analysing information, reasoning through defined problems, generating outputs and, within appropriate controls, taking action.
AI could increasingly extend across research, trading, risk management, compliance, lending, customer service, reconciliation, software development and internal operations. Simply adding an AI application to one of these functions, however, does not make an institution AI-native. The decisive change occurs when the underlying process itself is redesigned around what AI can do.
AI could consequently become less visible as it becomes more deeply embedded. Employees may interact primarily with the outcome of an AI-enabled workflow rather than consciously operating an AI tool. In reconciliation, for example, a system could process records, identify discrepancies, generate logs and assist with exceptions while the employee sees the completed workflow and focuses on judgement. In analysis and reporting, AI could gather and interpret information before presenting the resulting intelligence for human review.
Customers are likely to care even less about the technology itself. They want financial services that are accurate, timely, explainable and trustworthy. Whether AI assessed a transaction, analysed a document or detected suspicious activity matters less than whether the transaction was completed correctly, legitimate activity was not unnecessarily blocked and sensitive information remained secure.
The model itself may therefore become only one component of competitive advantage. As frontier models become more accessible, differentiation could increasingly come from proprietary data, institutional knowledge, distribution, customer relationships, regulatory understanding and the ability to redesign processes around AI. A powerful model does not possess the customer relationship, regulatory legitimacy or institutional context necessary to turn intelligence into a trusted financial product.
Enterprise adoption already points towards this organisational shift. Cognizant has been deploying AI to as many as 350,000 employees, while Access Bank has been testing engineering applications and IndusInd Bank has been experimenting with AI for knowledge-intensive work. Their importance lies not simply in user numbers but in the institutional learning generated when AI moves from individual experimentation into everyday organisational processes.
India provides an unusually consequential environment for this transition. It combines a vast consumer base, extensive banking networks, a major technology and services sector and digital public infrastructure operating at population scale, while presenting extraordinary linguistic, economic and geographical diversity.
The scope extends to work aimed at AI proficiency across more than ten of India's largest spoken languages, with applications extending beyond finance into agriculture, education, healthcare and skilling. For financial services, this could affect onboarding, financial education, customer service, credit interactions, fraud detection and communication with communities having very different levels of familiarity with formal finance.
The opportunity is increasingly central to Anthropic's India strategy. The company appointed Ghose as Managing Director for India in January 2026, ahead of opening its first Indian office. She brought more than three decades of experience scaling technology businesses, including enterprise AI adoption across banking and financial services, healthcare, manufacturing and government during her tenure as Managing Director of Microsoft India.
Anthropic has described India as its second-largest market globally for Claude.ai, while nearly half of Claude usage in India has been associated with computer and mathematical tasks, indicating substantial use for technically demanding work. Its Bengaluru presence forms part of a wider strategy involving enterprise, education and agriculture partnerships, with Ghose describing India as “very core” to Anthropic's global strategy.
Finance is particularly important because it combines scale, regulation, sensitive information, complex legacy technology and potentially severe consequences when systems fail. AI handling financial information therefore cannot be evaluated only through speed or cost. Accuracy, privacy, security, explainability, accountability, regulatory compliance and business continuity must be considered alongside capability.
Data privacy and data residency are particularly important requirements for enterprise AI in regulated sectors and privacy, security and governance need to be built into AI systems rather than added after deployment. In financial services, governance consequently becomes part of an institution's capacity to use advanced AI.
The organisational challenge is equally significant. Ghose has urged business leaders to use AI themselves and begin with the outcome they want rather than a catalogue of what the technology can do. An outcome-first organisation begins with a constraint and asks whether AI can remove it.
For a bank, that could mean improving customer service or risk management. An insurer might focus on underwriting or claims, an asset manager on research and information synthesis, and a fintech on fraud, customer acquisition or the economics of serving large populations. The technology becomes relevant after the problem has been identified.
This also changes how AI investment should be measured. Faster processing and lower operational costs are relatively easy to quantify, while the value of a new product, customer segment or previously unsolved problem is harder to capture through conventional productivity metrics. Falling frontier-AI costs could nevertheless make increasingly difficult and open-ended problems economically viable.
This does not imply complete automation. Finance involves decisions with significant consequences for individuals and markets and operates within substantial regulatory and accountability structures. The more plausible direction is a redistribution of work in which AI undertakes increasingly complex analytical and operational tasks while humans retain responsibility for supervision, judgement, exceptions and decisions where accountability cannot simply be delegated to a machine.
The competitive structure could consequently change. If AI reduces the cost of specialised knowledge, accelerates software development and embeds sophisticated analysis into ordinary workflows, advantages based purely on manpower or routine expertise may become less defensible. Proprietary data, customer trust, distribution, institutional knowledge and regulatory relationships could become more important. The AI race in financial services may therefore be less a race between models than a race to construct the strongest business system around increasingly accessible models.
India's position becomes especially consequential because it combines a major financial market, one of the world's largest technology-services ecosystems and digital public infrastructure operating at extraordinary scale. UPI and Aadhaar have demonstrated how digital infrastructure can support population-scale financial and identity services, while India's linguistic and socioeconomic diversity creates a demanding environment for AI deployment.
The transition is already entering the payments ecosystem. NPCI is developing a registry intended to verify and monitor AI agents conducting transactions on UPI under its emerging Unified Agentic Protocol, with early applications expected to focus on small, frequent payments before potentially expanding towards conditional purchases or investments.
That represents a movement from AI assisting people with financial activity towards AI potentially acting for them, making authentication, authorisation, monitoring, liability and governance central to future financial-market architecture. An AI agent capable of initiating a financial transaction is not simply a more sophisticated chatbot because it becomes an actor within the financial system.
The deeper proposition emerging during the GFF conversation is that AI's greatest financial advantage may not lie in doing today's work marginally faster, but in making previously impractical work economically and technically possible. That could mean understanding decades of legacy banking code, extending specialised knowledge across a broader workforce, solving complex analytical problems, developing new financial services or reaching customer communities that existing delivery models struggle to serve.
Efficiency creates the capacity, but the larger economic question is what an institution chooses to do with that capacity. The financial institution of the future may therefore not be the one in which AI is most visible. It could be the institution in which AI is embedded across research, technology, operations, risk management, customer service, compliance and decision-making until the technology itself becomes almost invisible, leaving behind its consequences - faster decisions, more accurate analysis, better products, systems understood sooner, previously inaccessible services made available and problems once considered too difficult finally solved.
That is the more consequential meaning of AI Native Markets. The question is no longer whether AI will enter financial markets because it already has, but what those markets become when intelligence is built into their underlying processes rather than attached to them as another tool.
At the GFF 2026 session, the arguments ultimately challenged financial institutions to look beyond the arithmetic of productivity towards the harder measure of technological value. That means, not merely how much work AI can take away from an organisation, but how much more the organisation becomes capable of doing because AI is there.