Creating a GenAI skilled workforce requires understanding of their personas. Firstly, we discovered GenAI usage varies depending on the role. For instance, we developed “experimental pods,” with folks with varying levels of expertise. This led to the development of persona-based use cases became primal, where GenAI tools are tailored to specific developer needs.

Secondly, the importance of high-quality data became evident. “Garbage in, Garbage out” applies to GenAI as well – clean data is essential for accurate and reliable results. Finally, it is crucial to understand the need for a cultural shift. GenAI isn’t just a new tool; it represents a whole new way of approaching software development. We had to help people understand that to work with GenAI, we must rewire the way in which we operate.

When working with clients, we take a similar approach: Rewire your software ecosystem (people, skills, and systems); clean data and choosing the right use cases aligned to the persona.

Digital Ascender, our GenAI tool was developed keeping these in mind. By empowering how our engineers work with GenAI, we also were able to develop a tool that makes sense for our clients.

Ascendion’s GenAI implementation has yielded impressive results. The Ascendion engineering experts working on AAVA platform team saw a significant boost in efficiency with the introduction of GenAI-powered unit test case generation (UTCG). For ex, engineers who code using Python and had to test a UTC with J-unit or C-sharp, saved as much as 45% in time and improved code quality and reduced defect leakage. GenAI further helps engineers generate better use cases leading to better problem solving.

Ascendion’s Digital Ascender (DA), a GenAI-powered tool for generating code solutions, proved invaluable for developers by providing multiple options and clear logic explanations. With DA and AAVA, we have accelerated development processes while maintaining high-quality outputs.

After months of honing our GenAI platform, we are now delivering solutions with GenAI-powered codebase.

GenAI centric solutions are driven by GenAI, making the entire work complementary work ecosystem. This in turn impacts speed to market, tremendously improves product quality, and most importantly, reduces technical debt!

Organizations are approaching AI in banking through two lenses.

The first is how AI can elevate day-to-day business operations, freeing talent from monotonous work so they can focus on more meaningful work and stay engaged. The second is how AI can accelerate technology to support the business, meaning the engineering side of the house

On the operations side, the use cases come directly from the teams living with the problem. A customer reaches the phone line and waits on hold for twenty minutes for a basic question. Can the AI handle standard FAQ responses automatically, so wait times drop, and the queue opens up for questions that need human depth? Password resets that need immediate resolution are another clear candidate. Account reconciliation involves huge volumes of numbers tied directly to the credibility of the bank, and to individual customers as well. If a customer needs to make an urgent payment but is locked out of their account, and the system cannot resolve it, that is a failure. The technology is not serving the business.

Until the model matures, the customer should always be able to reach a human. Standard questions, FAQs, and password resets are well within what a well-instrumented AI can handle reliably. Questions that need depth, context, or judgment are where a human needs to be available. In any regulated environment, such as banking and financial services, a human review is a mandate.

Alongside that, there needs to be a constant upgrade of context in the AI: a system or process in place to monitor real-time questions, keep FAQs current, and steadily expand what the AI can answer well. The crucial metrics that matter here are answer relevance, context relevance, and faithfulness. Additionally, without appropriate guardrails in place, models become vulnerable quickly, and in a banking environment, the consequences reach across security, credibility, and financial loss. Is the AI staying within the scope it has been set up and approved to cover? Monitoring the context consistently is what stabilizes adoption in production.

When AI adoption becomes operational, we can measure success in impact. What is the call wait time now? Are customers having their questions answered correctly with no misdirection? Has frustration gone down?

Ensuring that engineers unlearn, rewire, and get accustomed to the new ways of working is our priority. It’s a task to ensure that we are consciously looking at GenAI adoption at every phase in SDLC, be it for innovation or BAU execution.

Client considerations: Data, selection, and ethics

Clients need to understand three main aspects while embracing GenAI:

• Data readiness – data that is clean and of high-quality determines success
• Choose the right use cases that are best solved using GenAI
• Hype cycle will be around for a few years – understanding where they need to improve is essential, rather than one-size fits all approach

For ex, some problems can be solved through automation, some through Machine Learning algorithms, some through simple platform modernization. Clients need to understand that not everything requires a GenAI solution.

Ascendion works closely with clients to identify the optimal applications of GenAI, ensuring they leverage its capabilities effectively.
Security and ethics are also critical considerations. Understanding which data to use with specific generative AI models (is it proprietary, is it safe to feed data etc.), and language that is fed is vital. Ascendion has robust policies and practices in place to ensure ethical and compliant GenAI usage.

We train Ascenders to be cognizant of these ethical aspects both within our framework of operations and the client’s.

Ascendion’s GenAI journey is just getting started. We are constantly exploring new possibilities and applications. The focus is not on the specific models themselves, but on the art of prompt engineering and invoking any models as the situation demands. This empowers developers to unlock GenAI’s full potential. Challenges remain, but the potential benefits are undeniable. “We are currently in – well versed with art of possibilities – and exploring newer, farther frontiers.”