In this episode of GitHub for Leaders we look at how Banco de Crédito BCP planned and executed their transition to an AI-centric development model for their 1,400 devs in the highly regulated banking industry – a move which ultimately saved their organization 127K hrs over the last year.
Anjuan: Right now, AI is at the center of most breakthroughs in software development. And putting this new speed and scale in the hands of developers means that CXOs have the power to transform how software is built, tested, deployed, and secured. But having this much power means you also have to know how to maintain control, stay compliant, and understand the long term impact of these tools. And that���s what we���re going to look at today. To explain how senior leaders can navigate how to strategically select and then use AI tools as part of the software development process, I���m joined by two leaders from Banco de Cr��dito in Lima, Peru, the CIO, Maria Luisa Polo and Erika Le��n-Ravinez, DevSecOps leader. It���s so good to have both of you here with us. Tell me about the role AI plays in your organization software development strategy, and how did you get the buy in from around the company to implement this so swiftly? Maria: We started quite early in 2023 using generative AI with GitHub Copilot. So we were one of the first banks to decide to utilize this technology, and we faced issues of ignorance, of fear that was associated with past experiences that had occurred. With ChatGPT, for example, they didn���t necessarily have the security set in place that would be mature enough to utilize its technology in a bank. Anjuan: Now, one thing I was amazed to read was that you got 1,400 developers to use GitHub Copilot. How did you onboard them to using Copilot and how did you help them be successful? Erika: With GitHub Copilot, we started working on the entire coding front. The creation and the maintenance of the code itself. And this showed us that the engineer could actually dedicate 40% of their day to coding. And today we���re using, for example, 32% of all coding with generative AI. And this has obviously freed many of our hours across the bank. For example, this past year, we generated 127,000 free hours. Anjuan: You all had early pushback from people about using AI in a bank that has so many concerns about risk and security. And I���m wondering how have the people that were early skeptics about these tools changed because you���ve gotten the result of having engineers being more productive, more efficient, and faster to get products out their door. How has their view of AI changed in the face of those results? Erika: What I can say is that today, the engineers are really happy using the tool. We have an NPS score of 70, and 83% of them feel more productive than they did before. They say the quality of their code has improved with the tool. 99% have a positive perception of using the tool, and honestly, they���re really satisfied. And on the C-level side, there���s continued investment in AI as part of the BCP. This productivity speaks for itself. We���re freeing up a lot of hours every year, and we want to complete as much of our software development journey as we can with generative AI. Right now, 31% of our journey is supported by AI, and we aim to reach 49% by the end of the year and expect even more the next year. Maria: And to respond a bit to your question about cybersecurity and risk, I think they���re just as strict as they were at the beginning. But I think this experience has helped us identify the requirements and conditions that any initiative needs to meet under Credicorp���s AI program. And now, what we���ve done is create a clearer path so that future initiatives know exactly what they need to comply with to move forward and develop. Anjuan: Now, I was amazed to read that last year you freed up 32% of coding time for your developers, and that you have a goal of hitting 30% next year, but you���ve already saved 127,000 hours of developer time. And I���m super curious, what are your developers doing with all that extra time? Maria: Look, we operate in two dimensions. The first dimension produces more, and, in fact, we have metrics that show this. There���s a lot more delivery, and the unit cost per delivery is getting lower every time. Let���s say that, with those two metrics, we can demonstrate these freed up hours have been used to boost productivity, or be more productive. Anjuan: Now, you two work at a bank and banks are in a highly regulated industry with lots of compliance, lots of security. And one of the things that leadership had to do was make the case for using GitHub Copilot in this industry. What are some of the things that leadership did to get Copilot in the doors at the bank? Erika: Like Maria was saying, in the beginning, there wasn���t much knowledge, and there wasn���t really talent with AI or tech specialization. And as one of the first banks in Peru using AI technology, we had no point of reference, no one to compare ourselves to. And security was key, it always has been, but at that time, the fear of data loss, the fear of information leaving the company. And cybersecurity and risk departments were just starting to gain recognition when it came to this technology. Still, the team always had a clear goal that we wanted to use AI in our day-to-day work. GitHub helped us train in different areas, and not just in tech. This support made a huge difference, this partnership helped us a lot. Most of all, it helped us move forward and convince everyone that this was the best solution. Anjuan: Now you two are at the forefront of financial institutions that are using AI to realize tremendous productivity gains from your engineers. And I���m sure that you���re hearing from your peers questions about how do we get started? So when people ask you those questions, what are the first one or two things that you tell them to do before they start their own AI journey? Maria: I���d say first of all, you have to try it. This is a technology you need to test. You don���t have to overthink it or wait for the perfect use case with huge productivity promises to incorporate this into your process or day to day. Second, and we already mentioned this, is having trained people. You have to invest in both training and technological adoption of this technology. And third, and I���d say this is the most important, is your board has to believe this is good for the bank. Anjuan: GitHub Copilot has evolved from a simple AI coding assistant to an AI collaborator that your organization can put to work right now. This is a moment in time when senior leaders can transform the way their organization works and make a major impact on how their software is built, improved, and secured. And you can learn more about how to take those next steps at this link.
