By Alex Pazik, Sr. Technical Account Manager (ISV) – AWS
By Utkarsh Contractor, Field CTO – Aisera
Aisera
Artificial intelligence (AI) and Large Language Models (LLMs) are driving efficiency gains across the financial services and banking industry. Advances in LLMs, including the development of domain-specific LLMs tailored for specific industries, have created new possibilities for integrating with AI applications in various fields. Gartner predicts that by 2027, over 50% of enterprise large language models will be domain-specific, marking a 49% increase from 2023.
Aisera provides customers proprietary, domain, and industry-specific LLMs that deliver precise, context-aware insights and intelligence. Using Amazon Elastic Compute Cloud (EC2) and P5 instance types, Aisera is able to support fine-tuning these LLMs with AWS’s GPU-based instances. With Aisera’s AI Agent Platform, powered by Amazon Web Services (AWS), organizations can integrate prebuilt AI agents and Amazon Bedrock Agents to manage complex workflows through multi-agent collaboration.
Evaluating AI Agents
Aisera recently announced a new benchmarking framework, CLASSic, for evaluating the performance of agents in real-world enterprise applications. This framework provides a holistic approach to evaluating enterprise agents across Cost, Latency, Accuracy, Stability, and Security. The study that produced the CLASSic framework found domain-specific agents outperformed frontier LLM-built agents, demonstrating the advantages of domain specialization in enterprise applications. The study also found that in complex domains like IT, FinTech, and HR, Aisera’s agents outperformed general-purpose LLMs with 76.1% accuracy and 99.3% success rate in refusing harmful prompts. And at $0.005 per request, Aisera’s agents can help customers save up to 80% on AI processing costs while maintaining competitive response times and accuracy.
Aisera’s AI Stack
Depicted in Figure 1, Aisera’s universal, domain, and task agents leverage a combination of Large Foundation Models (LFMs), domain-specific LLMs, and task-specific Small Language Models (SLMs). This combination provides customers’ accuracy at 1/10th the cost of using general-purpose LLMs. LFMs are a broader category that includes LLMs, and other models trained on various types of data (like images, audio, or video) for general-purpose tasks. SLMs are models of natural language processing (NLP) optimized for efficiency with a lightweight architecture that requires less computational power and memory.
Figure 1: System of AI agents with universal agents, domain agents, and task agents.
Depicted in Figure 2, Aisera’s AI Agent Platform offers pre-built, out-of-the-box agents that automate high-volume, repetitive tasks such as customer inquiries, unauthorized activity detection alerts, and compliance-related documentation. These agents, which integrate with existing financial platforms such as Zendesk, Genesys, NICE, and SAP, expand Aisera’s ability to support a comprehensive range of AI driven use cases:
Human-like customer support: Provide real-time, personalized customer service through chatbots.
Account setup and management: Assist customers with authentication, account setup, and registration.
Payment and deposit tracking: Notify customers of upcoming payment due dates, accessing account information securely to verify transactions and process payments.
Unauthorized and anomalous activity detection: Detect anomalies and flag potential unauthorized activity in real time.
Automated document processing: Extract insights from contracts, regulatory filings, and reports.
Figure 2: Enterprise AI architecture diagram showing the integration of Aisera’s AI Agent Platform with enterprise systems.
Case Study
Launched in 2017 and headquartered in Los Angeles, Dave offers a free mobile checking account and numerous industry-first features like no-interest overdraft alternatives, free credit-building through rent payment reporting. With over 5 million users, Dave had to scale support efficiently with maintaining response and customer satisfaction score (CSAT) targets.
DaveGPT, an AI assistant built on Aisera’s AI Platform, provides real-time information and personalized support options to customers that have helped increase member satisfaction and retention. With sub-5 second retrieval-augmented generation (RAG) powered by Amazon OpenSearch and dynamic workflow agents, Dave achieved 89% chatbot resolution and 60% first-call resolution (FCR). Dave’s partnership with Aisera to leverage agentic AI supports the neobank’s continued innovation in the industry.
“Today’s consumer expects personalized assistance on demand. Partnering with Aisera has enabled us to address our members’ needs in real-time through AI. DaveGPT, powered by Aisera, fundamentally impacts how we scale by automating simple interactions and generating value through efficiency and supporting our members.” – Jason Wilk, CEO and founder of Dave.
Conclusion
The advancement of AI and domain-specific LLMs has created opportunities for transformation across the financial services and banking industry. Aisera’s AI Agent Platform, powered by AWS, demonstrates the capabilities of these domain-specific LLMs to deliver accurate, cost-effective AI solutions for this sector.
Next Steps
Contact Aisera to learn how their domain-specific AI solutions built on AWS can transform customer service and employee support across various types of industries including Financial Services. Visit AiseraPlay to get hands-on experience with AI agents and explore how they can transform your IT, HR, Customer Support, and Financial Operations or schedule a demo today.
.
Aisera – AWS Partner Spotlight
Aisera is redefining how organizations operate and deliver value to customers leveraging AI to automate service requests, simplify operations, and improve customer satisfaction across industries. Aisera is recognized by AWS for their Conversational AI and Generative AI Software Competency.
Contact Aisera | Partner Overview | AWS Marketplace