DANA’s Sasono: “Differentiate between GenAI hype and fundamentals”

Generative AI (GenAI) innovations have created waves throughout the financial industry, with the potential to impact numerous processes across the value chain. Many financial institutions (Fis) are exploring new applications primarily targeted at improving internal processes and raising productivity.

DANA, Indonesia’s leading digital wallet and payment company serving 180 million users, has integrated multiple use cases for AI and GenAI in particular to boost productivity and enhance services. Norman Sasono, CTO of DANA, shared the company’s strategy around AI, current use cases, and challenges the company faces in implementation.

Building an AI strategy

The launch of ChatGPT and other large language models (LLMs) has led to a spate of Gen AI use cases across the industry and democratised its access. Compared with traditional AI, Gen AI has a superior ability to generate new content from users’ natural language requests.

Sasono said: "Al is no longer a technology that is accessible to just a certain few, but it is now available for everyone, both individual at personal level and business at organisational level. [But] we need to really understand what it actually is able to do; separate the hype from the fundamentals with the euphoria about the ChatGPT, LLM, and GenAl."  

He argued that GenAl is ideal for things like content generation, software code creation, conversational user interfaces like chatbots, and knowledge management. However, for areas like prediction and forecasting, decision intelligence, segmentation, and recommendation engine personalisation, traditional AI and machine learning (ML) capabilities are better suited.  

Emerging use cases for AI

DANA has embraced this wave of innovation by launching internal AI initiatives, particularly in strengthening operations.

Sasono said: "As we build some use cases, we already see the fruits in terms of productivity gains and advances, efficiency gains, and competitive advantage"

"We have utilised Al models for personalisation through Al-based tailored recommendations. We are also using Al for credit scoring, fraud detection, scam scoring, and internal operations monitoring, also using it for augmented software development."  

For GenAI, it is still early days. DANA’s main use cases with Gen AI have been in AI-augmented software development with GitHub Copilot, the AI-enabled DIANA Chatbot for customer service, product development, and areas like document creation. The company is also seeing an impact on software development efficiency.

“Since January this year, all our engineers have been using GitHub Copilot to assist them in augmenting software development with AI. We saw dramatic results with up to 50,000 lines of code generated by AI every month,” said Sasono. With this, the company integrated with partners faster, saving 90% of development time.

Emerging challenges in scaling AI

To apply AI at scale, Fis must have a strategic roadmap for enterprise-wide operationalisation with the right technology and data foundation.

“The real competitive value of AI is if your company has a rich, comprehensive data set. Without valuable data, AI may not necessarily be transformative,” Sasono emphasised.

The model quality and its ability to give accurate and consistent results are among key concerns. FIs must ensure that they have unbiased models, wider data sets with assurance of fairness, and transparency. Another significant challenge is the availability of skilled personnel and the leadership buy-in.

“You need a team that has the capability to execute this, along with the vision and direction from leadership. It cannot be sporadic departmental-level efforts to be transformative,” said Sasono.

He elaborated that evolving regulatory compliance and ethical considerations, require robust governance frameworks to mitigate risks. FIs must also address emerging risks from potential exploitation of AI by malicious actors. For instance, developments like deep fakes require upgrades to authentication and security measures.

DANA is adopting a strategy to infuse broader AI adoption across its processes and to optimise its transformative potential through various use cases.

However, the journey is not without challenges for any institution. Keeping pace with rapidly evolving technology and industry developments, ensuring the right data quality, and using the right tools to prevent AI hallucinations, inconsistencies, and biases remain some of the foremost concerns. Companies must also navigate evolving regulatory landscapes and address concerns like vendor concentration risks and data security risks.

Building scale in AI adoption will require a strategic approach  to enable a holistic transformation through process restructuring, talent acquisition and reskilling, and managing an organisation’s cultural and mindset shift.