GitHub Copilot Accelerates Coding at Saxo Bank, Unblocking Developers and Enabling Smarter Debugging

Saxo Bank’s founding vision in 1992 was to make global financial markets accessible to more people. For decades, it has pioneered the use of technology to this end. In 1998, Saxo was among the first in Europe to launch an online trading platform, providing professional investment tools to a new segment of clients. Today, the broker serves over 1.3 million clients, collaborates with more than 150 partner banks and brokers, and handles 200 million API requests every day. This technological prowess positions Saxo as a leader in adopting cutting-edge innovations like AI. 

What sets Saxo apart is not just its use of AI for specific applications but its commitment to a comprehensive, organization-wide AI strategy. This approach supports its mission to “Get More Curious People Invested in the World”, while also enhancing its risk management and operational efficiency. “Everything we're doing with AI, or any other technology, is closely aligned with our overall objectives and business plan for 2025,” says Mattias Movin, Saxo’s Head of Group Enterprise Architecture. “We're not doing AI for the sake of AI; it is always to deliver on things like increasing client and employee satisfaction, reducing costs, reducing risks and increasing revenue.” 

With that in mind, Saxo wants to consider AI for use cases in all areas, including technology, workplace, business processes and client-facing capabilities. Its strategy rests on five key pillars: identifying use cases, adopting AI responsibly, building a foundation of quality data, building a culture of AI literacy, and ensuring AI frameworks are adaptable. To implement the strategy, Saxo has adopted numerous technologies from Microsoft, which has worked closely with the company to help it use AI in a more agile way. “We have benefitted a lot from the fact that Microsoft was so fast in making OpenAI models available in Azure, and luckily we were able to get early access,” says Jeppe Reitz, Saxo’s Global Head of Data Science. “Without that we wouldn't have been able to do what we’ve done,” he adds. “You can imagine the restrictions involved in a financial institution like ours, and the vetting process to make these services available. But we had the chance to get the technology within our own network and gain these functionalities very quickly after they started getting popular.”  

Microsoft played a key role in helping to identify use cases: it hosted three ideation workshops at Microsoft’s offices in Denmark. Two product teams, both responsible for a different client journey, attended each workshop. After Saxo presented its strategy, Microsoft's experts explained what its AI services had to offer, before facilitating brainstorming sessions about how to implement them at Saxo. 

Coding is now faster, easier and more accessible 

One tool that has proved especially transformational at Saxo is GitHub Copilot, the AI coding assistant for developers. Saxo built its digital trading platform itself and has a team of more than 700 developers who maintain and update it. More than 400 of them have already started using GitHub Copilot since it was introduced in February 2023. 

“It might not have been around for long, but it's just part of the toolbox now and is used in the creation of almost all new code,” says Peder Holdgaard Pedersen, Principal Developer at Saxo. “People expect to have this tool now.” The bank’s developers typically use GitHub Copilot to generate short lines of code, which they then review. The team’s overall acceptance rate for the code is around 30%. “It speeds up a lot of work because it unblocks our developers,” Pedersen says.  

It has also made coding more accessible for team members who are not trained developers, he adds. “I know that some non-development staff or development adjacent staff use it, such as infrastructure engineers who don't write much code but still get a lot of benefit from having AI that is tailored towards technical use, code and development.” For developers, it has made it possible to engage with languages and frameworks they are not familiar with. “It’s also useful if you have a bug and need to figure out why the application or service is malfunctioning,” Pedersen says. “You have the error messages, and you have the code, and you work with AI to identify the problem.”   

One team member used the tool to write a script that performed a rote task he had long disliked doing, Pedersen adds. The tool also has a feature that is similar to predictive text on a word processor: “It predicts what you're going to type, and it gets better at predicting it all the time.”  Each developer uses the tool in different ways, he adds. “I might use it to create a lot of test code rapidly, or someone else might use it if they’re stuck on a problem, and work with AI as a sort of sparring partner. Others create boilerplate code with it.”  

Change management for an AI literate workforce 

Saxo’s team is working towards making AI opportunity mapping an integrated part of the design phase of any new service. To do this, it aims to ingrain AI in its culture by enlisting employees to scout for new use cases in client journeys. But before choosing to use any AI technology, it is essential to identify a business need, says Tania Cocesiu, Product Owner for Microsoft 365 at Saxo. “It's important to have a clear understanding of any goals to be able to deliver outcomes.”  

Making sure that use cases demonstrate clear business benefits is one way she is driving widespread adoption of Microsoft 365 Copilot at the organization. “We have offered licenses to everyone in Saxo – almost 2,600 people,” she says. “The main driver for this decision was that we were looking at how innovative technologies can benefit our employees in terms of productivity, quality of work, and wellbeing.” 

“Everyone got the chance to explore the tool and assess how it can assist them, and we've been quite satisfied with this inclusive approach,” she says. “We want to upskill the whole workforce so everyone can play a role in driving innovation with these technologies,” adds Movin. 

Saxo measures the success of the tools by employees' perceptions of whether they are saving time and doing good work, and the figures so far from an internal employee survey are encouraging. It found that people believe that they save time with Copilot, are more productive and able to complete tasks faster. Overall, employees found Copilot 365 useful and said they would like to keep using it in the future.  

“The results were much better than we expected,” Cocesiu says. “It bodes well for the future of generative AI at Saxo; people are eager to use these technologies.” Such enthusiastic adoption is possible because fostering AI literacy and changing the culture is a cornerstone of Saxo’s AI strategy. “The implementation of this shouldn't be seen as a technology project, but a change management project,” Cocesiu says. “Team members need to be educated and made curious about using AI tools–it's not something that just happens on its own.” 

AI champions and Center of excellence help move strategy forward  

As well as communicating the benefits of AI tools on the company’s internal channels, Cocesiu’s team also organizes live demos, which are presented by employees who make up Saxo’s network of AI Champions. 

These 80 employees were selected by their leaders because they showed a particular enthusiasm for adopting the new technology. They hailed from different offices across the globe, so that Saxo could gain a cross functional and global perspective on the changes it was making. “I have seen employees incorporating Copilot into their daily routines for tasks such as catching up on emails and preparing for meetings. Impressively, we have also identified use cases where employees use it for developing business cases, marketing campaigns, and content creation,” Cocesiu says. 

Another key part of driving adoption is getting leaders on board, she adds. “We've got a lot of support from leaders in encouraging their employees and making Microsoft 365 Copilot visible.” Underpinning all of this is Saxo’s new AI Center of Excellence, a team whose role is to focus on enabling various AI capabilities. This involves maintaining a backlog of incoming AI ideas, consulting and supporting different initiatives, encouraging a consistent approach to the technology while pooling skills, and clustering use cases with high value and the potential for reuse.  “The Center of Excellence delivers competencies, tools and frameworks to meet the needs of different business lines as they develop use cases,” Movin says.  

At the same time, various steering committees are ensuring use cases are responsible and ethical. Saxo already has robust governance in place to manage risk around data and machine learning. It is now building on that to reflect the fact that generative AI requires additional frameworks to ensure it is used responsibly.  Members of groups such as the AI Center of Excellence and Model Risk Management develop these frameworks based on several principles. One is that staff members should only use AI tools, models and services that are approved by Saxo, to ensure they are legally compliant and to reduce risks.  

Another principle is that employees are accountable for AI-generated content, meaning they must always validate it thoroughly before using or distributing it. They must also disclose the use of AI in their work to maintain trust with colleagues and clients. “Being responsible for security, privacy, ethical standards and legal compliance helps to form the foundation for identifying use cases,” Movin says.  

Enhancing client Satisfaction with Client-Facing AI Solutions  

Saxo has also developed a chatbot on Azure OpenAI that offers its clients faster service.  

Because of its generative AI component, the answers feel much more tailored and relevant, but the bot does not have free reign to reply in the same way another Large Language Model would. “We're a very professional broker, so we don't want to let it loose,” says Nicolas Palm Perez, Saxo’s Head of Predictive Modelling and AI. “We ensure that we're in full control of what answers the chatbot can give, and the help pages that it can refer clients to.”  

This bot, and other potential use cases, are helping to make the arcane world of investing more accessible to clients who don’t necessarily have specialist knowledge. Capturing that segment of the market will help to drive growth long term, says Reitz, the Global Head of Data Science. “Back in the day, people in the financial domain using services like ours were professionals,” he says. “Now, retail investors are the biggest growing segment in this domain, and they need more help because they're not spending 24 hours a day navigating what happens in the market.” 

Saxo is considering multiple ways to offer its clients more. “We want to go more in-depth in terms of portfolio overviews and giving clients clarity around their financial situation, and what they can expect in the future,” Reitz says. “That will involve much more guidance, helping them navigate and interface our applications in using common language.” 

Saxo is also working on a news recommended feed engine that will help drive traffic and engagement, attracting new clients and encouraging existing ones to deepen their relationship with Saxo. To speed up time to market for this use case, Saxo recently sent a group of its engineers to work on two prototypes using Large Language Models at Microsoft’s AI Co-Innovation Lab in Munich—a cutting-edge facility where teams get tailored guidance for their projects. 

The recommender engine puts Saxo at the forefront of an emerging trend, Reitz says, but soon enough he expects services like these to become the norm. “Historically you have had tons of information, tons of instruments, tons of ways to look at price data, and now this can be easily accessed,” he says. “The way you get prices and financial news can be in ordinary colloquial language, as well as an overview of risk on your portfolio. It will change the way people navigate financial data and get their financial overview, that's certain. And we need to be able to cater for that.”