Inside Money20/20: The Hard Reality of AI Adoption in Finance

The Money 20/20 conference is one of the most influential gatherings shaping the future of finance, payments, and data-driven innovation. Our team had the chance to join Money20/20 USA in Las Vegas this October, bringing back some insightful takeaways (more on that below).

Bill Sanders, Business Development Leader at Janea Systems, joined thousands of senior executives on the ground. And you know, what? Instead of asking what AI can do, leading financial organizations now focus on building trust, strong infrastructure, and delivering tangible results. As Bill notes,

The era of experiments is ending. The era of enterprise-grade execution has begun.

In this recap, we unpack the defining ideas, frameworks, and case studies from Money20/20 in 2025. Continue reading and you will find out how leading financial institutions turn AI from a concept into a competitive advantage for the years ahead.

AI Maturity Curve: From Crawl-Walk-Run to Cultural Transformation

At Money20/20, the conversation around AI maturity took on a new tone. It’s no longer just a technical milestone, but a cultural one. Banking leaders realize that building models is the easy part. Scaling, maintaining, and integrating them across an enterprise is where the real work begins.

The Crawl-Walk-Run Model

Sumee Seetharaman, Head of AI/ML at TD Bank, shared a new perspective on this journey: crawl, walk, run.

• Crawl (The Past Decade): Banks spent years experimenting with predictive AI, including automation, marketing optimization, and credit adjudication. It was about using data to make better decisions in existing systems.
• Walk (The Present): Now, they tackle generative AI. Seetharaman explained how TD Bank spent months just getting the foundational model stack right, followed by another 6 to 9 months building the engineering infrastructure needed to move from proof of concept to production.
• Run (The Future): TD Bank is now gearing up for the next leap — agentic AI. Even for an industry leader, this remains uncharted territory.

This transition from experiments with predictive AI to building the generative AI infrastructure is a common failure point for many organizations. A mature rapid prototyping strategy is essential to bridge this gap successfully.

The Day-Two Rigor

Seetharaman pointed out that the hardest part begins after deployment. Models, she said, “deprecate really, really fast.” Customer behaviors shift, data changes, and expectations evolve.

To handle this, banks need what she calls Day-Two Rigor: a disciplined process to maintain and refresh AI systems over time. It requires not only technical scaffolding (monitoring, retraining, explainability) but procedural consistency.

The Day-Two Rigor effectively turns the platform into a powerful engine for enterprise AI knowledge management, allowing thousands of employees to build their own tools. In short, AI maturity means treating models like living systems, not one-time projects.

For many institutions, this rigor is complicated by the need to bridge legacy systems with modern cloud infrastructure, a foundational challenge in scaling AI.

Case Study in AI Maturity: BNY

While TD Bank focused on the technical journey, BNY CEO Robin Vince tackled the cultural side. In his keynote, Vince reframed AI maturity as a human challenge:

It’s not about computing power; it’s about adoption, culture, and inertia.

BNY’s approach centers on Eliza, an in-house AI platform designed to democratize access across the company. This strategy effectively turns the platform into a powerful engine for enterprise AI knowledge management.

Vince believes that “AI is for everyone, everywhere, everything at BNY.” The numbers back it up:

• Over 15,000 employees have used Eliza to build their own AI tools.
• 1,000+ employees have completed over 40 hours of AI training.
• The bank has 150 AI-powered solutions in active development.

This mass participation model turns AI into a cultural engine. It creates what Vince calls a “flywheel of excitement” — a feedback loop where curiosity drives experimentation, which in turn drives adoption. Chances are, enabling small-scale AI “quick wins” is a viable strategy for driving enterprise-wide adoption.

The New Definition of AI Maturity

The conference’s discussions revealed that true AI maturity blends technical rigor with cultural engagement.

A perfect model without adoption dies in isolation. A culture that loves innovation but lacks discipline breeds chaos. The most advanced banks try to balance both — to build strong technical scaffolds while nurturing a workforce ready to use, question, and improve them.

Operationalizing Intelligence: The AI Factory

The "AI in Finance" panel, featuring NVIDIA, Stripe, TD Bank, and Fiserv, offered a metaphor for operationalizing AI at scale: the AI Factory. Why a factory?  scaling artificial intelligence from one-off experiments to a repeatable, industrial-grade capability that drives real business outcomes.

Three Layers of Scalable AI

The panelists outlined a framework built on three interconnected layers, each playing a role in how organizations operationalize intelligence.

Data Ingestion Layer

This foundational layer captures proprietary data from every possible touchpoint — taps, swipes, logins, and account openings. It’s the company’s most defensible asset and the bedrock of every intelligent system that follows.

Inference Layer

Here’s where heavy lifting happens. The Inference Layer is home to the organization’s proprietary foundation models that analyze patterns, learn from behavior, and generate insights. It’s designed for scale and reliability, ensuring that intelligence is consistent across the business.

Decentralized Agents

These are the specialized AI tools that go into production. Each agent, whether it’s a checkout assistant, a fraud detector, or a customer service bot, is powered by the central Inference Layer. Together, they form a network of intelligent apps that continuously learn and improve.

Case Studies: Proprietary Foundation Models

The AI Factory concept explains why many financial institutions invest in their own foundation models instead of relying on third-party APIs.

• Stripe: Josh Ackerman shared that Stripe built what he called the world’s first payments foundation model. This core Inference Layer was explicitly engineered to outperform older fraud detection systems that could no longer keep pace with evolving threats.
• TD Bank: Sumee Seetharaman confirmed – they developed their own predictive model called PRISM. Trained on the bank’s proprietary Data Ingestion Layer, PRISM was able to detect subtle customer patterns that off-the-shelf models missed, dramatically improving personalization and accuracy.

The takeaway: proprietary data is the real differentiator. Generic models trained on public datasets can’t capture the nuances unique to a company’s customers. To derive that hidden value, organizations need to train their own foundation models on their own data.

Agentic AI and the Headaches That Come with It

Agentic AI – autonomous systems capable of acting on behalf of users, from making purchases to handling payments – took the center stage at Money 20/20. AI agents promise a radical shift in how consumers and businesses interact. On the flip side, it creates new kinds of friction that the industry has barely begun to address.

Agents as the New Middlemen

During the AI in Finance panel, Josh Ackerman described agentic commerce as “the killer use case” for AI-driven transactions, though he admitted the field is still in its infancy.

• PayPal and Google Cloud announced a partnership to develop agent-driven shopping experiences that merge Google’s Conversational Commerce agent with PayPal’s payment systems.
• Stripe is collaborating with OpenAI on an instant checkout feature that allows Etsy or Shopify sellers to process transactions directly through ChatGPT.
• American Express CTO Hilary Packer led a session titled “From Act to Transact: Empowering AI Agents to Pay,” signaling a fundamental strategic buy-in from one of the world's largest payment networks.

The Merchant's Dilemma: Mastercard

Not everyone is swept up in excitement. Johan Gerber, EVP of Security Solutions at Mastercard, reminded that agentic commerce may introduce as many risks as it solves.

1. Security and Fraud. When an AI agent acts as a proxy for a user, fraud systems lose visibility into the individual behind the transaction. Traditional trust signals (device ID, geolocation, behavior) disappear, breaking the models that keep commerce safe.
2. Authentication. Strong customer authentication doesn’t map neatly to a world where an agent makes decisions on behalf of a person. As Gerber put it, “How does a merchant know the agent is real and acting with true consent?”
3. Marketing and Commoditization. If agents choose products purely on data and price, brand loyalty becomes meaningless. Merchants face a new dilemma: how do you convince an algorithm to pick your product over an identical competitor’s? The risk is a race to the bottom where the lowest price always wins.

Building the Architecture of Trust

The panelists agreed on one point: the trillion-dollar potential of agentic commerce depends on creating a new foundation of trust. The old principle of Know Your Customer (KYC) must evolve into Know Your Agent (KYA) — a framework where businesses can verify that an agent is legitimate and truly authorized to act on a user’s behalf.

Several companies are already working on this next layer of trust:

• Stripe is developing shared payment tokens to create a consistent trust mechanism across the financial ecosystem.
• Mastercard has teamed up with Cloudflare to define security, privacy, and data governance standards that AI agents must follow.
• Checkout.com is positioning itself as an integrator, connecting everyone in the agentic ecosystem, from large language model providers to card networks like Visa and Mastercard.

These challenges are not side issues — they’re the admission price for the next era of digital commerce. The promise of agentic AI will remain limited until the industry solves the KYA problem. Behind the scenes, a new competition is emerging over who will set the standards of agentic trust.

Strategic Recommendations for Financial Services Leaders

Money20/20 2025 marked a turning point for the financial sector. The conversation shifted from exploring AI’s potential to executing on it – a hard-line focus on operational readiness we also observed at the AI4 2025 conference.

Here are the top 4 takeaways and strategic recommendations for decision-makers in the finance industry:

1. Build Your Own AI Factory Blueprint

Buying off-the-shelf AI tools is no longer enough. The future belongs to institutions creating scalable and standardized AI factories built on proprietary data.

Start by asking:

• What does our Data Ingestion Layer look like?
• Do we treat our proprietary data as a competitive advantage?
• Do we have a credible plan to build our own Inference Layer?

And remember: depending on third-party models that can’t access your core data means limiting your strategic control over AI implementation outcomes.

2. Move from KYC to KYA

Google, PayPal, Amex, and Stripe deploy autonomous agents that buy, recommend, and transact on behalf of users. This shift introduces a new challenge: trust. Fraud detection, authentication, and transparency are all being redefined. Every player in the payments ecosystem must establish a Know Your Agent (KYA) strategy.

Institutions have three paths forward:

• Collaborate on new standards like Mastercard and Cloudflare.
• Develop trust infrastructure like Stripe’s shared payment tokens.
• Or risk being “blinded” by agent-driven transactions they can’t verify.

3. Master Day-Two Rigor to Drive ROI

Once the models are deployed, the hard part begins. Executives need to pivot investment toward training and integration, not just model creation.

Ask yourself:

• Who owns Day-Two Rigor — the continuous process of keeping AI relevant and reliable?
• How are we equipping all the employees to use AI tools, not just the 15 who built them?

ROI follows when the entire organization becomes fluent in applying AI to everyday work.

4. Define Your Role in the Agentic Ecosystem

Agentic AI is the next distribution channel, and every company must decide where it fits.

Are you a:

• Platform creating agent experiences, like Google or PayPal?
• Partner integrating those agents into your ecosystem, like Etsy or Shopify?
• Enabler building the trust and payment rails that make it all possible, like Stripe or Mastercard?

Your relevance in the coming decade will depend on how clearly you define your position in this emerging architecture of trust.

---

If your organization is ready to move beyond the AI strategy and start building scalable AI systems, this is the time to act.

At Janea Systems, we help financial and data leaders design their own AI Factory through our AI Maturity Workshops. These hands-on sessions assess your current capabilities, uncover critical gaps, and chart a clear path toward enterprise-scale AI adoption.

Let’s turn your AI ambition into a measurable impact. Get in touch with us using the contact form below.