Solutions & Use Cases

Five Questions Banks Must Ask Before Choosing AI

Kognitos

Five Key Questions for Banks and Financial Institutions to Uncover the Best AI Amid Growing Skepticism

Key Takeaways

Written by Kognitos CEO Binny Gill, this post addresses growing skepticism about AI’s ROI in banking and finance, citing Gartner’s forecast that many generative AI projects stall after proof-of-concept and Wall Street’s doubts, yet argues the technology is a genuine game-changer for complex financial workflows like Procure-to-Pay, Order-to-Cash, and Record-to-Report. Its core contribution is five questions institutions should ask when evaluating AI vendors: can they explain automated processes in plain language; can they customize LLMs and cloud environments; do processes include human oversight and verification; do they offer both factual (deterministic) and intuitive (generative) capabilities; and will they use your data to train models outside your organization. The post champions smaller, specialized LLMs over general-purpose ones and stresses transparency, accuracy, and data privacy. The takeaway: cut through the hype by scrutinizing vendors against these criteria. Explore the Kognitos platform.

How financial leaders can navigate the rush to adopt AI and find the ideal solution for their business.

By Binny Gill, CEO of Kognitos

Enterprises across every industry are eagerly jumping on the AI bandwagon, driven by the promise of unparalleled efficiency, innovation, and a supposed competitive edge. However, the journey from ambition to real-world implementation is fraught with significant challenges, especially for the banking and finance sectors.

While businesses initially embraced AI with enthusiasm, there is now growing skepticism about the tangible ROI that AI can deliver. Major media outlets are questioning why seven leading tech giants are doubting the technology’s long-term investment viability; while others are asking, “Has the AI bubble burst?” Some hedge funds have even warned investors to be skeptical of companies like Nvidia, while others suggest Big Tech is struggling to convince Wall Street that AI investments will bring real returns altogether.

Recent insights from Gartner underscore these challenges, predicting that 30 percent of generative AI projects will be abandoned after the proof-of-concept stage by 2025. Major financial institutions like Goldman Sachs echoed this cautionary stance, recently releasing a report downplaying the so-called “AI gold rush,” describing the promised ROI from Silicon Valley as little more than snake oil, a sentiment shared by Barclays and Sequoia Capital.

So, what’s the verdict? Is AI just another overhyped trend destined to fade away? Not quite. There’s more to the story than the doubters suggest.

At the enterprise level, scaling AI solutions, ensuring security and ethical compliance, and managing increasing costs, particularly those associated with training large language models (LLMs), present challenges. But the release of OpenAI’s GPT-4o mini has reignited discussions on the long-term viability of AI adoption, spotlighting a shift towards smaller, specialized LLMs.

Are these specialized AIs more valuable than general-purpose ones? As companies navigate AI’s vast potential, many remain unsure of the most effective use cases, often realizing they don’t know what they don’t know.

For financial leaders, the potential benefits of generative AI extend beyond the hype. Financial processes that are integral across organizations, like Procure to Pay (P2P), Order to Cash (O2C), and Record to Report (R2R), can gain significantly from these advanced capabilities. While some may be skeptical of yet another automation promise, it is essential to take a holistic view.

Embracing AI’s potential streamlines workflows, fosters innovation, and helps maintain a competitive edge in a rapidly evolving market. For financial leaders, this all begs one major question: How can we make it work for our business?

How Financial Institutions Should Evaluate AI Providers

Effective evaluation of smaller AI solutions requires asking the right questions. By zeroing in on these crucial inquiries, organizations can meticulously assess AI models and vendors and thoroughly address concerns about the safety and efficacy of AI technologies. This approach ensures that the solutions they choose are not only trustworthy but also perfectly tailored to their specific needs and risk profiles.

Here are the right questions to ask:

Can the vendor clearly explain automated processes in plain language?

Ensure the AI vendor provides straightforward descriptions of all automated processes. This transparency helps stakeholders understand the AI system, verify compliance with standards, and build trust in the vendor’s accountability.

Can the vendor customize LLMs and cloud environments for your needs?

Confirm the vendor’s ability to tailor LLMs and cloud setups to your specific requirements. Customization enhances performance, security, and compliance, aligning the AI solution with your strategic goals.

Do AI processes include human oversight and verification?

Check if the AI system allows for human control and review. This is essential for ensuring accuracy, reliability, and ethical use, helping to prevent errors and biases while maintaining system integrity.

Does the vendor offer both factual and intuitive AI capabilities?

Ensure the vendor supports both deterministic (fact-based) and generative (intuitive) AI processes. This combination leverages accuracy and creativity, enhancing decision-making and operational efficiency.

Does the vendor use your data to train models outside your organization?

Confirm if the vendor uses your data for training models beyond your control. Protecting your data ensures privacy, safeguards intellectual property, and maintains compliance with data protection regulations.

The Next Frontier for Financial Workflow Transformation

Traditional automation has hit a wall when it comes to the complexity of financial workflows. But, without all the hyperbole, AI is a real game-changer. It can handle tasks previously deemed impossible, turbocharging productivity and slashing costs across financial operations. Unlike rigid, high-maintenance predecessors, AI adapts, learns, and evolves.

Unlike earlier automation tools, AI’s adaptability and learning capabilities allow it to handle intricate, cross-functional processes. This flexibility, coupled with generative AI’s broad applications, positions AI as a transformative technology for the financial industry. Business leaders just need to know how to implement it.

Imagine workflows that seamlessly connect your entire organization, unlocking hidden value. Yes, there are challenges, transparency, accuracy, and privacy always are, but with careful scrutiny, these can be managed. By addressing these challenges and asking the right questions, financial institutions can unlock new opportunities to streamline operations, drive innovation, and maintain a competitive edge.

Frequently Asked Questions

Banks should ask: (1) Can the vendor clearly explain automated processes in plain language? (2) Can the vendor customize LLMs and cloud environments for your specific needs? (3) Do AI processes include human oversight and verification? (4) Does the vendor offer both factual (deterministic) and intuitive (generative) AI capabilities? (5) Does the vendor use your data to train models outside your organization? These five questions help financial institutions evaluate AI vendors on transparency, customization, oversight, capability breadth, and data privacy before committing to a platform.

Major financial institutions like Goldman Sachs, Barclays, and Sequoia Capital have raised doubts about AI's promised returns, describing some Silicon Valley claims as overblown. Gartner predicted that 30 percent of generative AI projects will be abandoned after the proof-of-concept stage by 2025. Challenges include scaling AI securely, managing rising costs for training large language models, and identifying the most effective enterprise use cases. Despite these concerns, AI still holds genuine transformative potential for financial workflows when implemented with proper scrutiny.

Traditional automation tools are rigid and high-maintenance, often hitting a wall when confronted with the complexity of financial workflows. AI, by contrast, adapts, learns, and evolves to handle intricate cross-functional processes. Generative AI's broad applicability allows it to turbocharge productivity and slash costs in areas like Procure to Pay, Order to Cash, and Record to Report. This adaptability positions AI as a genuinely transformative technology rather than just another iteration of rules-based automation.

Human oversight ensures that AI-driven processes remain accurate, reliable, and ethically sound by allowing staff to review and control automated decisions. Without it, errors and biases can propagate undetected through financial operations, creating compliance risks and potential regulatory exposure. Financial institutions should confirm that any AI system they adopt provides mechanisms for human verification and intervention. This oversight is especially critical in highly regulated industries where accountability and auditability are non-negotiable requirements.

Deterministic AI, sometimes called factual or rules-based AI, produces consistent, verifiable outputs based on defined logic, making it ideal for processes that require precision and auditability. Generative AI introduces intuitive, creative capabilities that can handle unstructured data and novel scenarios where rigid rules fall short. For financial institutions, the best AI platforms support both modes, combining the accuracy required for compliance-sensitive tasks with the flexibility needed to handle complex, exception-heavy workflows. Vendors that offer only one mode may leave gaps in operational coverage.

Financial leaders should directly ask vendors whether they use customer data to train AI models that operate outside the customer's organizational control. Using proprietary financial data to train shared or external models creates risks around data privacy, intellectual property protection, and regulatory compliance. Vendors should be able to provide clear contractual commitments that data remains within the institution's controlled environment. This evaluation is especially important given strict data protection regulations and the sensitive nature of financial transaction data.

Core financial processes that span entire organizations stand to gain the most from generative AI, including Procure to Pay (P2P), Order to Cash (O2C), and Record to Report (R2R). These workflows are complex, cross-functional, and often laden with exceptions that traditional automation cannot handle well. AI can streamline these end-to-end processes, reduce manual intervention, lower costs, and improve accuracy at scale. By automating previously intractable tasks, financial institutions can unlock hidden operational value and maintain a competitive edge in a rapidly evolving market.

Kognitos