At a glance
Assistive technologies enabler supporting accessibility to banking services, leveraging Google Vertex AI
Solution category
Technology
Geography
Global
Built by
GFT / Google Cloud / Thought Machine
The challenge
Banks have a high number of staff performing online bank assistant duties. A lot of these activities are redundant and repetitive. There is a need to increase the efficiency of customer service teams by freeing them from routine call or chat handling to focus on tasks that add more value.
Banks also need to drive operational cost savings, which will allow them to reinvest on transformation-led initiatives, enabling them to drive customer acquisition and retention.
There is an increasing focus to significantly improve the customer experience compared to standard AI based chatbots, delivering more personalised and trusted support.
The solution
The GenAI Intelligent Assistant has been implemented with:
- A front end implemented in React.
- A gateway and authentication system using JSON Web Tokens (JWT).
- An orchestration engine running on Google Kubernetes
- Engine (GKE) that manages the interaction of the system with the LLMs.
- An integration layer that manages streamed updates from the core banking system by managing a transaction store, allowing access to a single source of truth for data.
Benefits
1. Reduced operating costs by providing customers with a new way of requesting services and managing their finances.
2. Improved customer engagement and satisfaction with improved net promoter score (NPS) ratings.
3. Quicker access to services and improved overall service performance.
4. Improved employee experience – reduction of repetitive tasks; focus on higher value tasks; change of role from policy implementer to troubleshooter.
.jpg)
How the solution works
The GenAI Intelligent Assistant extracts value from the capabilities of LLMs in the following ways:
- Zero-shot learning (prompting): decodes the intent of the user query using a call to a LLM.
- Category prediction: LLMs have been used to generalise or specialise categories to determine item relevance.
- Query type inference: LLMs have been used to determine what type of query will produce the answer the user is looking for.
- Inference: the user is able to ask questions, and the system is able to understand the intent, and answer accordingly - showing evidence from their account (e.g. money spent in coffee shops)
Once information satisfying the user’s request has been extracted, it is pushed into the LLM to generate a response.
Using a natural language interface creates an expectation from the user that an interactive and ongoing conversation with the system will be possible.
Security and reliability
A proprietary decoder / encoder solution has been developed that first ensures an untrusted prompts can’t be used to create unexpected behaviour and only trusted data is used to generate a response for the user. Secondly, the solution uses data based verification to remove ‘hallucinations’, outputs that the neural nets have simply invented.
Proactivity and personalisation
Users expect personalisation from their conversations with an agent. They expect agents to remember their previous interactions and to draw on information shared. Additionally, interactions with other users can be used to educate the system on intents - to better understand what a user might be asking. In this sense, the assistant can help users ask questions and discover information that they would never have thought to look for without it.
"
The IVN approach allows us to combine Google Cloud's scalable infrastructure and cutting-edge technologies like GenAI, with GFT's deep domain expertise and Thought Machine's next-generation core banking platform. This powerful synergy enables us to deliver transformative solutions that address the complex challenges of core banking modernisation, from legacy system migration to real-time data processing
"
Anil Saboo
Director ISV Partnerships
GCP
Why the partnership?
Thought Machine is becoming the de facto option for tier 1 banks undertaking core transformation, as well as smaller banks and fintech challengers who are redefining the industry.
Google is driving the art of the possible with GenAI through their offerings including Gemini, Vertex AI, and PaLM API. These are value-added services based on Google Cloud technologies through which banks can leverage Gen AI and LLMs.
GFT is the preferred integration partner for many clients worldwide, utilising expertise in innovation, AI and specialist engineering skills. As part of GFT’s AI.DA marketplace, there are a number of use cases that banks can build on top of the Gen AI Intelligent Assistant.
Together, through this offering, Thought Machine, Google and GFT can deliver significant value, cost efficiencies and innovation for banks.
No items found.