When good insights lead to confusing experiences
Banks know a lot about their customers. They know how often someone logs in, what products they use, how they tend to spend or save, and even what they might need next. That level of insight has never been the problem. The problem is what happens next.
A customer might receive an email encouraging them to open a savings account. At the same time, they might see a message in their banking app about a credit product. If they call support, the conversation might go in a different direction. Each interaction makes sense on its own. But together, they feel disconnected.
From the customer’s point of view, the bank is not acting like one organization. It feels like a collection of systems making independent decisions. That is the challenge this article focuses on.
From Hackathon idea to real-world use case: Decision-led banking
This use case was originally developed by the Banco de Bogotá team as part of the 2025 SAS Hackathon. Their goal was to explore how banks can move from fragmented customer engagement to a more coordinated, decision-led approach—where decisions are made once and carried through consistently across channels.
The approach itself is straightforward. The bank started with basic information, such as demographics and preferences, to build an initial understanding. It then learned from customer behavior over time through transactions and interactions. As that picture became clearer, it continuously refined how customers were grouped and used that understanding to guide what the bank should say or do next.
This type of approach is often referred to as customer 360 or customer segmentation. It works well for understanding customers. But it still leaves a key question unanswered: How does the bank turn that understanding into one clear, consistent action?
Why consistency is so hard
In most banks, decisions are not made in one place. Marketing, digital, service, and product teams often manage their own channels, tools, and priorities. Even when they are working from the same customer data, they are not always acting on it in the same way. This is why customers receive mixed messages.
It is not a data problem. It is a coordination problem. And most banks are not structured to solve that problem centrally.
A different approach: decide once, act everywhere
Instead of letting every system decide what to do independently, a better approach is to make one decision and use it everywhere. That decision answers a simple question: What is the single most relevant action for this customer right now, and how do we make sure every channel delivers that same message?
This is where SAS Intelligent Decisioning comes into play. It allows banks to bring together data, models, and business rules in one place so decisions can be made consistently.
But there is still another challenge. Even when the bank knows the right action, it must explain why that action was chosen, communicate it clearly to the customer, and make sure the tone fits the situation.
This is where agentic AI becomes useful. It allows banks to support decision-making with AI in a controlled way, improving how decisions are communicated and understood without changing how they are made.
What an AI agent actually does in this context
An AI agent in banking is not a chatbot, and it is not replacing decisions. It acts more like a coordinator within the decisioning process, helping translate decisions into clear, consistent customer interactions.
It helps with two important things:
- Turning decisions into clear communication
- Turning technical reasoning into understandable explanations
For example, a model might indicate that a customer is likely to benefit from a savings product, while business rules confirm they are eligible and that the timing makes sense. The decisioning system selects that action, and the AI agent helps translate it into a message the customer can understand.
At that point, the AI agent helps draft a message that is clear and appropriate, translates technical reasoning into simple language, and ensures the communication aligns with the bank’s tone and policies.
The decision remains controlled and governed. The AI helps make that decision easier to understand and easier to deliver.
Where the SAS Agentic AI Accelerator fits in
Up to this point, the focus has been on making better decisions and keeping those decisions consistent. The next step is making those decisions easier to communicate and easier to understand. This is where the SAS Agentic AI Accelerator fits in.
The accelerator provides a structured way to introduce AI into existing decision workflows without disrupting how decisions are made. Instead of treating AI as a separate experiment, it becomes part of the system in a controlled and repeatable way.
In this example, the accelerator would be used to build an AI agent that sits alongside the decisioning process and helps with two key tasks.
First, it helps turn decisions into clear communication. Once a decision is made, the agent can generate customer-facing messages that are consistent in tone, aligned with policy, and tailored to the situation.
Second, it helps turn technical logic into understandable explanations. It can take model outputs and decision rules and translate them into simple language that can be used by customer support teams or even shared directly with customers when appropriate.
The important part is that the AI is not making the decision—it is supporting how that decision is delivered. By using the Agentic AI Accelerator, teams can build this capability once and reuse it across channels, instead of relying on separate teams or systems to interpret decisions in different ways.
Bringing it all together
The Banco de Bogotá use case shows how banks can better understand their customers over time. The next step is making sure that understanding leads to clear, consistent action.
By combining customer insight and segmentation, centralized decisioning, and agentic AI for communication and explanation, banks can move toward a model where customers receive fewer but more relevant messages, interactions feel more intentional, and decisions are easier to apply consistently and explain clearly.
This is not about adding more technology. It is about making existing capabilities work together more effectively.
Customers do not see systems, models, or channels. They experience conversations.
When those conversations are consistent, trust grows. When they are inconsistent, even strong insights lose their value. Agentic AI, used in the right way, helps bridge that gap—not by making decisions on its own, but by helping organizations act on their decisions more clearly and consistently.
This is where the Agentic AI Accelerator becomes practical. It provides the structure to build these agent-based capabilities in a way that is consistent, reusable, and aligned with how decisions are already managed in SAS Viya.
Explore the Agentic AI Accelerator
If this approach resonates, a good next step is to explore the SAS Agentic AI Accelerator on GitHub. The repository provides a practical starting point for building agent-based workflows in SAS Viya, including examples, reusable components, and guidance on how to bring together decisioning, AI, and orchestration in a structured way.
It is designed to help teams move beyond ideas and experiments and begin building real, governed solutions that can scale. Whether you are just starting to explore agentic AI or looking to expand existing decisioning workflows, it offers a clear path to begin testing, learning, and applying these concepts in your own environment.
From Hackathon idea to real-world use case
If you are interested in how this idea was developed, you can explore the team and their work on the SAS Hackathon platform: Banco de Bogotá – SAS Hackers Hub.
The SAS Hackathon provides a space for teams like this to take real business challenges, combine data, analytics, and AI, and build practical solutions in a short period of time. Many of the concepts explored in this article, including centralized decisioning and agentic AI, are being actively tested and refined through these hackathon projects.
This is a strong example of how innovation does not always start with large transformation programs. Sometimes it starts with a focused use case, a small team, and the right tools to connect insight to action.
