Unlocking Generative AI in Finance

The AI revolution is here – or so we’re told. Yet, while the technology’s potential is undeniable, most enterprises seem stuck in pilot mode, unable to push GenAI beyond controlled experiments. Across sectors, adoption has been largely confined to trials and proof of value experiments. Even Accenture, the global consulting powerhouse, reported that the bulk of its over one billion US dollars in GenAI projects came from pilots, not full-scale implementations.

Yet there are notable outliers: pioneering financial institutions that are safely leveraging generative AI at scale. They enhance decision-making with insights from enterprise data, streamline operations by automating workflows, and grow revenues by accelerating investment due diligence and risk assessment. At Squirro, we’ve had a front-row seat to their success, powered by our enterprise GenAI platform.

In this article, we explore key lessons we learned while solving complex data challenges for central banks and government authorities – lessons that played a crucial role in helping us unlock the true value of generative AI in this demanding sector.

A Sprawling GenAI Provider Landscape

GenAI truly went mainstream with the public launch of ChatGPT in 2022. Since then, a cottage industry of GenAI providers has emerged, targeting specific use cases and audiences. Meanwhile, industry giants like Microsoft and Intel have introduced their offerings, catering to an eager market. IDC, a global market research, analysis, and advisory firm, expects spending on GenAI to outpace the broader AI market, with a projected five-year CAGR of 59.2%, reaching $202 billion by 2028.

BFSI's Cautious Approach

But within the banking, financial services, and insurance sector, GenAI’s transformational potential has been tempered by real-world risks. The sector’s reliance on personal data and its harsh regulations have led to heightened expectations of CISOs and their peers:

● Privacy: Financial institutions are obliged to protect customer data at all costs, as data breaches can lead to significant financial losses, reputational damage, and legal penalties.

● Security: In BFSI, the integrity of all elements of the IT system is critical. A single vulnerability could compromise entire systems, with potentially ruinous consequences.

● Accuracy: Financial decisions are grounded in rigorous data analysis. Interpretation errors can lead to misguided strategies, poor execution, and financial losses.

● Flexibility: GenAI deployments are rarely greenfield operations but need to integrate into existing infrastructure. This requires versatile, LLM-agnostic platforms that connect easily with existing data sources.

● Scalability: Financial institutions deal with massive volumes of data. GenAI solutions need to be able to be scalable to meet growing data volumes and new use cases without compromising privacy, security, accuracy and cost-effectiveness.

The industry’s prudence has been well-founded. Checking all of these boxes while delivering workable solutions on time and to budget has proven to be such a formidable challenge that, as of now, only few GenAI providers have come close.

Squirro’s Journey From Enterprise Search to GenAI

As the broader industry raced to catch up with GenAI, Squirro had a head start, having spent the decade preceding its public launch diligently solving what has since become the core challenge of enterprise-grade AI: accurate, secure, and permissions-enabled enterprise search.

Each successive project – carried out with leading governmental regulators, national banks, and financial service providers – further refined our offering, ironing out wrinkles in search optimization, information retrieval, and, crucially, permission-enabled access control.

When GenAI took off, our mature information retrieval stack, already adept at finding the most relevant enterprise data to answer user queries, paired perfectly with the newly available large language models. Combined, they enabled advanced retrieval augmented generation, which has since become the industry standard for enterprise GenAI solutions.

The result, today, is an offering that perfectly aligns with the BFSI sector's stringent requirements for privacy, security, and accuracy at scale. Today, our success in breaking into the financial sector is evidenced by partnerships with the European Central Bank (ECB), Deutsche Bundesbank (DBB), Bank of England (BoE), and Standard Chartered Bank.

Streamlining Investment Analysis and Due Diligence

To illustrate our impact, consider our collaboration with a development bank financing private-sector projects promoting economic growth and sustainable development in Latin America. Using Squirro, the bank can seamlessly aggregate relevant data from internal and external data sources and automate the drafting of entire sections of the investment proposals. By minimizing manual effort and speeding up the process, Squirro not only lowers the cost per opportunity but also enhances the quantity and quality of closed deals.

This type of use case can be replicated endlessly across organizations to streamline employee onboarding, automate the extraction of ESG-related data from reports, contracts, and filings, accelerate product and market research, expedite RFP (request for proposal) responses, enable hyper-personalized customer experiences, and support legal document discovery.

The Relentless Pursuit For Accuracy, Security, and Privacy – at Scale

Our mature enterprise search and information retrieval offering served as a solid foundation for our ongoing pursuit to increase the accuracy, security, and privacy of RAG-based GenAI in production-scale deployments:

In 2024, we acquired Synaptica, a leading provider of an enterprise taxonomy, ontology, and knowledge graph management system to enhance the accuracy and reliability of AI-generated responses. The same year, we invested in Meetsynthia.ai, pioneers in AI guardrail management aimed at mitigating organizational AI risks, reducing legal and operational liabilities, and safeguarding against AI hallucinations. And we partnered with RAW Labs to support the integration of real-time operational data into GenAI use cases.

Meanwhile, we developed an agent framework to streamline the creation, deployment, and management of AI agents that can interact with tools to create, analyze, enrich, classify, fact-check, and transform data. Underlying these efforts, we continue to enhance our solution’s security and privacy by adhering to industry standards, enabling organizations to control their data with robust safeguards for sensitive information.

Driving the AI Revolution

The result is a future-proof, enterprise-grade GenAI platform that empowers banks, financial institutions, and other large organizations to harness the full potential of their data. And this is just the beginning! By continuing to innovate and refine our offerings, we aim to be at the forefront of the ongoing GenAI revolution enabling organizations to operate with unparalleled efficiency, accuracy, and security.

Learn more at squirro.com