How AI is changing the game for banks: a discussion with Microsoft

Kathleen Woodard has a unique perspective on the value of agentic artificial intelligence (AI) in banking. After spending 26 years as a banker, she now serves as Microsoft's Head of Banking, Industry Advisory, Americas, working closely with the largest financial institutions in the region.

We recently sat down together to discuss the evolution of AI in banking, the importance of a platform-based approach, and how banks can accelerate AI adoption — as well as the risks of not doing so. Kathleen even shared her top tips for what senior banking leaders can do today to not get left behind by the AI wave.

You can check out the highlights here or watch the full recording of episode 29 of the Banking Reinvented podcast below for all the insights.

Let's start with the proportion of banks that are already developing or implementing generative AI. According to Bain, 59% of banks surveyed are already exploring generative AI for customer service use cases, while 55% are exploring it for software code development.

But how did banks first experiment with generative AI? Well, it all started with tools like GitHub Copilot that can generate code, but banks adopted them cautiously, implementing strict guardrails and focusing solely on internal use cases. After that, they began to launch chatbots for simple customer queries.

And now, with the introduction of agentic AI, we're finally entering the next stage in banking — one that will have a massive impact on the industry.

Initially, many banks thought of AI mainly as a way to improve productivity at their call centers. But now, with the power of autonomous AI agents, it's become clear that the impact can be so much greater, whether it's by reducing labor costs, providing personalized customer experiences, or generating higher revenue through cross-sell, according to Microsoft.

Of course, banks would do well to take a human-in-the-loop approach, but AI agents can already help simplify banking operations, accelerate value generation, and reduce admin burdens at multiple stages across end-to-end journeys and workflows. Why waste hours of time scanning documents for information when an AI agent can do this in a matter of minutes?

However, getting the most out of AI agents can be difficult — if not impossible — if you don't have an engagement banking platform in place. These platforms make it easy to add an agentic architecture layer so you can integrate AI agents throughout your departments and processes.

As Kathleen told me, Microsoft is always "customer zero" for its own innovations, so when they developed Copilot, their internal teams immediately started experimenting with this new tool. The results speak for themselves:

• An 8% increase in the number of sales leads generated
• 40% faster close rates on sales deals
• On track to deliver cost savings of $400M across all divisions

Of course, banks will have different success metrics and may be slower to implement AI tools like this, but it's clear that using AI to address simple customer service queries is just scratching the surface of what's possible.

Many banks are wondering whether they should buy existing agentic AI tools or build their own. The answer is a bit of both.

Once your bank has an engagement banking platform in place, you can start with out-of-the-box capabilities to get your AI agents up and running. Then, using the same platform, your developers can build the tools and workflows that are unique to your organization, helping you deliver the experiences that will set you apart.

With everything available on a single platform, AI agents can access the accurate data and capabilities they need in real time. Take mortgage lending as an example. AI agents can collect documents, validate the information provided, review creditworthiness, and connect all stakeholders on the platform.

So far, we've taken a global perspective, but how are U.S. banks in particular approaching agentic AI?

Well, banks throughout the U.S. are operating in a very tough regulatory environment that's constantly evolving as new AI developments emerge. Combine that with legacy systems and it's easy to see why the adoption of new tech can be slow and expensive throughout the region.

However, things are already looking up. Most U.S. banks have had chatbots for years, but more and more are looking to transition to AI customer services agents, instead. This is mainly due to improvements in conversational AI tech, which eliminates the cold, generic chatbot responses that often frustrate customers.

And while AI agents aren't able to fully replace human interactions, they're already helping U.S. banks quickly resolve simple customer queries in a friendly, personalized way — covering everything from explaining bank statements to selecting the right credit card for their needs.

Early adopters of AI agents are already starting to see the benefits, but don't forget that, inevitably, there are always laggards in the adoption cycle. These laggards may not admit to experiencing any obvious downsides yet, but the gap between them and the leading banks will rapidly widen in the next few years.

In fact, if your bank isn't already experimenting with agentic AI, then you're already falling behind, as Backbase Head of AI Chris Shayan explained in a recent episode of the podcast. And even if you're experimenting, it will become increasingly hard to catch up to the frontrunners. But it's not too late.

By developing a solid agentic AI strategy and implementing a platform approach, your bank can prepare to take advantage of agentic AI developments for the benefit of both your customers and your internal teams. Better yet, costs and implementation timeframes have been significantly reduced in recent months, so it's the perfect time to take the leap.

Even though the potential of agentic AI is clear, it's quite possible you'll encounter some resistance from your internal teams who are worried it will replace their jobs or trigger a loss of customer interactions. You may even be sure how your bank should approach agentic AI in the first place.

Katherine's advice is simple: "Get comfortable with the technology yourself," she said. Then, you can empower your teams to experiment with it, including the best ways to use it in their day-to-day workflows.

It's already possible for your bank to implement AI, and in the years to come, it will quickly become an imperative. But don't forget, there are a few things you'll need to keep in mind beforehand:

1. Shifting your organization's mindset — leaders need to encourage a positive, open-minded approach to AI across all employees. Teams must start looking at how agentic AI can create real value for them and their customers, rather than fearing it.
2. Adopting an engagement banking platform — having a unified platform in place will help to connect all your stakeholders, data, and tools, allowing you to leverage an agentic architecture. Better yet, it will make it easy to perform updates and release new products, as needed.
3. Harnessing progressive modernization — we recommend dreaming big about the potential of agentic AI at your bank, but start small with implementing it. By taking a progressive, step-by-step approach, you'll not only keep up with tech developments, you'll also continue to meet your customer's evolving expectations.

Check out the full episode of Banking Reinvented here, or check out previous episodes to learn more about the future of banking — and how your financial institution can prepare for it.