Inside Moody's Bottom-Up Approach to AI

INTERVIEW
Sergio Gago on Scaling AI at Moody’s

Sergio Gago, head of AI & Quantum Computing at Moody's, discusses how the financial services firm has identified hundreds of AI use cases by involving employees in the innovation process, moving beyond an initial approach that relied on a specialized innovation team.

In this “Five Minutes With” interview, Gago explains Moody’s approach to AI implementation, including the development of tools like a Policy Assistant that answers compliance questions and a Credit Memo Generator that automates financial analysis. He addresses why innovation teams often struggle to scale AI solutions when operating in isolation and shares practical insights for implementing AI in financial institutions.

By encouraging employees to contribute ideas, Moody’s generated over 800 potential AI applications. This inclusive strategy has helped the firm uncover numerous opportunities, such as automating compliance queries and generating credit memos.

Gago also explores the potential of augmented reality in the financial sector and offers advice for entrepreneurs building AI tools for finance.

This interview has been edited for clarity and length.

How has Moody's approach to AI evolved? 

The traditional approach to innovation in many organizations, including initially at Moody’s, often involves creating specialized innovation teams tasked with exploring cutting-edge technologies like AI. These teams work on a few use cases to demonstrate the potential of the technology. However, this approach has significant limitations, particularly when it comes to scaling AI solutions.

First, these innovation teams tend to operate in isolation, disconnected from the core business units. This can result in a lack of buy-in from the teams who ultimately need to adopt these innovations. If the business units don’t see value or if the solutions aren’t directly relevant to their challenges, these projects rarely progress beyond the proof-of-concept (PoC) stage.

Second, innovation teams often focus on flashy, “wow-factor” use cases that demonstrate the capabilities of AI but lack the practical feasibility to address real-world needs. For example, creating a chatbot or generating a credit report may look impressive in a demo, but issues like data quality, hallucinations, security, and compliance often prevent these solutions from being implemented at scale.

Additionally, in most organizations, legal, compliance, and cybersecurity teams are often brought into the process late. The earlier they get involved, the smoother the implementation process.

Finally, innovation teams can’t change company culture on their own. Successful AI implementation requires a broader organizational shift, where every employee feels empowered to explore and integrate AI into their workflows. This means providing training, creating open collaboration platforms, and fostering a mindset where innovation is everyone’s responsibility—not just the job of a small, siloed team.

At Moody’s, senior leaders called on employees to help shape the urgent evolution in GenAI and encouraged them to play an important role in identifying creative solutions and ways to unlock its potential. 

This open, bottom-up approach let all of our employees become prompt engineers if you will and generated over 800 ideas and secured organization-wide buy-in. It also allowed us to uncover those atomic use cases - the small 5-10 minute daily tasks - where AI can drive a lot of cumulative efficiency gains.

Tell me about some use of those use cases  

Policy Assistant: Answering Compliance Questions with High Accuracy

At Moody’s, we’re a large, regulated organization with rigorous compliance policies—covering everything from travel to client interactions. Employees often have detailed questions like: `Can I accept this gift from a client?’ or ‘Am I allowed to fly business class for this specific trip?’ Answering a high volume of questions required our compliance team to spend time and bandwidth, which they could easily devote to addressing more complex issues their job demands.

To address this, we built an internal Policy Assistant using AI. It processes and interprets policy documents, incorporating all recent updates. Employees can now ask the assistant directly and receive accurate, compliant answers almost instantly. We’ve designed it to mitigate risks significantly—especially for critical compliance issues. This tool not only improves efficiency but also allows our compliance team to focus on more complex, high-value tasks. 

EDFX Navigator Research Assistant: Providing Insights for External and Internal Users

The EDFX Navigator Research Assistant is perhaps one of our flagship AI-driven tools. It’s designed to sit on top of our extensive data platforms and acts as an intelligent assistant for both internal teams and our customers. It synthesizes vast amounts of data and presents actionable insights tailored to the user’s query.

Internally, our analysts use it to quickly extract key insights during client meetings or while preparing reports. 

Externally, it’s a powerful resource for our clients, helping them navigate market dynamics, identify risks and opportunities, and make informed decisions without having to comb through raw datasets themselves. What makes this tool special is its ability to provide contextualized insights—turning raw data into a narrative that’s immediately useful. It’s a game-changer for how we deliver value through our data.

Credit Memo Generator: Automating Financial Data Interpretation

The Credit Memo Generator is one of the standout tools we’ve developed. Traditionally, creating a credit memo—summarizing a company’s financial position and creditworthiness—was a manual, labor-intensive task. Analysts had to sift through financial statements, pull in data, and compile into a coherent report.

Using AI, we automated this process. The Credit Memo Generator extracts data from structured sources, analyzes it, and generates a detailed credit memo in minutes. It doesn’t just summarize numbers—it provides a narrative, highlighting key metrics, identifying potential risks, and even flagging anomalies or trends. This has drastically reduced the time it takes to produce these reports and ensures consistency and accuracy across the board. Analysts now have more time to focus on deeper insights and strategic decisions.

How is Moody’s exploring augmented reality (AR) in conjunction with AI?

We see immense potential in combining AI with AR to transform how employees interact with data and make decisions. Imagine an analyst visualizing complex datasets in 3D or a manager simulating economic scenarios within an immersive AR environment.

For entrepreneurs or investors looking at this space, what's your advice on how to approach AI in finance?

My key advice would be to always focus on real-world applications, not just novel research for research's sake. You absolutely need that deep domain knowledge to build something useful.

We see a lot of impressive academic work, but when you dig into it, it often doesn't reflect how financial institutions actually operate. What looks great in a paper may be impractical to implement.

So I'd say to engage practitioners as early as possible to pressure test your solution. Sit down with the intended end users to understand their pain points, their workflows, and the regulatory constraints they're under. That's how you'll identify those use cases where AI can drive real impact - by embedding it intelligently within existing processes.

The goal should be to empower the human experts with better tools and insights, not to replace them with some black box. That collaborative, human-in-the-loop approach is where I think you'll see the most progress in the near term.

What has surprised you so far?

I'd say the biggest surprise was the eagerness and speed with which the entire organization embraced AI, when given the right tools and context. People ran with it and found innovative ways to apply it to their day-to-day work. They quickly saw how it could be transformative.

Another thing that surprised me was how many other institutions - banks, asset managers, insurers, even regulators - have reached out to learn from our approach. They want to understand our governance model, and how we're deploying responsibly.

IN CASE YOU MISSED IT
Recent Five Minutes with Interviews

• Ravenpack | Bigdata.com’s Aakarsh Ramchandani on AI and NLPs

• PhD candidate Alex Kim on executive tone in earnings calls

• MDOTM’s Peter Zangari, PhD, on AI For Portfolio Management

• Arta’s Chirag Yagnik on AI-powered wealth management

• Finster’s Sid Jayakumar on AI agents for Wall Street

Thanks for reading!

Drop me a line if you have story ideas, research, or upcoming conferences to share. [email protected]