MoneySuperMarket’s AI leap with Firemind and AWS

By: Francisco Martins, Solutions Architect – AWS
By: Chris Owen, Chief Architect – MONY Group
By: Anna-Riitta Vuorenmaa, Chief Marketing Officer – Firemind

Firemind

MoneySuperMarket, AWS, and AWS Partner Firemind deployed an AI-powered assistant using Amazon Bedrock, moving from proof-of-concept (PoC) to production in just 6 months. While many organizations stall at the PoC stage when exploring generative AI, MONY Group plc (FTSE:MONY) and their brand MoneySuperMarket (MSM) took a different path, helping thousands of customers make smarter financial decisions faster with AWS.

Opportunity

MSM helps customers and households save money by simplifying search and comparison of deals across products such as home insurance, car insurance, travel insurance, credit cards, energy, and broadband.

They found that searching for the right credit card is often overwhelming, with customers having to evaluate introductory offers, understand if there’s an interest-free period, see if they meet the different eligibility criteria, and compare the product fees across different cards and providers. MSM decided to use AI to simplify this customer journey through the introduction of Money Concierge, an AI-powered assistant built on AWS.

Execution

To accelerate the path to production, MSM created a new team responsible for researching and testing AI technologies for their use cases, starting with a few members and recently expanding in size. The team was designed with the flexibility for experimentation, a strong sense of urgency, and with the autonomy to quickly identify and iterate on the PoCs. To accelerate the path to production, they chose Firemind as their AWS Partner, moving the Money Concierge solution from PoC to production within 6 months.

Solution architecture

MSM integrated Money Concierge into their credit card channel. When customers are evaluating different credit card offerings, they can select a widget that starts the Money Concierge experience. This assistant:

• Understands the user’s financial position by extracting insights from completed forms and their credit report, such as their annual income, current indebtedness, credit balance needs.
• Uses the third-party credit card providers’ APIs to get an updated overview of the available offers and terms.
• Presents personalized suggestions in natural language, considers the user’s credit balance context and suggests a sample repayment plan customers can use as guidance to clear debt before interest applies. Figure 1 illustrates the Money Concierge experience with a sample repayment plan.

Figure 1: Money Concierge Experience and sample repayment plan

When MSM and Firemind decided to build Money Concierge with AWS services, they had a strong preference for fully managed or serverless services. This was to remove the administrative burden of managing servers, allowing product teams to focus on building features that sustain a strong customer experience.

• Amazon Simple Storage Service (Amazon S3) is the main storage service for documentation and terms and conditions (T&C) of the different credit card offerings. It also hosts the static website as the origin for the content delivery network (CDN).
• AWS Lambda is the main serverless compute service used to run code without managing servers. It’s used first as an API endpoint to receive customer requests and starts an AWS Step Functions workflow. Lambda functions attach to the private subnets of MSM’s VPC, with no access to the internet, allowing the functions to access private API endpoints and securely retrieve data from credit reports and offers.
• AWS Step Functions orchestrates the different steps required to build the customer credit profile. It coordinates the different backend workflows that collect and enrich customer data, while tracking the workflow state across steps. Finally, it generates the custom prompts enriched with customer context that are going to be sent to Amazon Bedrock for the final customer output.
• Amazon Bedrock provides secure and scalable generative AI model access on-demand. VPC-attached Lambda functions connect privately to Amazon Bedrock through an AWS PrivateLink VPC endpoint, so that data doesn’t leave AWS and is kept under MSM’s control. Another key advantage of Amazon Bedrock is that data isn’t used to improve the underlying large language model (LLM) performance, adding another layer of data privacy in the context of responsible AI. Requests to the Amazon Bedrock API use a secure (TLS) connection, which encrypts data in transit. Amazon Bedrock Knowledge Bases embed key MSM documentation and T&C, delivering accurate context-aware generation through Retrieval-Augmented Generation (RAG). This feature uses Amazon OpenSearch Serverless as the underlying vector store.

Figure 2: AWS Architecture supporting Money Concierge

This solution uses a highly available and scalable modular architecture. Figure 2 illustrates the AWS Architecture and services supporting the Money Concierge solution. MSM deployed the workload across different Availability Zones within the selected AWS Region, and the underlying services automatically adjust capacity based on the actual workload demand. By designing for modularity, MSM can also extend the solution to additional financial products in the future.

Results

MSM’s Money Concierge has helped over 55,000 customers in their credit card journey. MSM’s customers receive:

• Faster access to relevant card options based on eligibility.
• Clearer explanations of card benefits and limitations.
• Greater confidence in their decision-making, leading to improved conversion rates.

Money Concierge provides improved customer engagement and conversion and received 93% positive feedback.

Conclusion

This post illustrates how AWS and partners like Firemind can help organizations accelerate their ambitions to deliver customer-facing AI solutions into production. The key items to drive value are:

• Start with the customer; identify their pain points and work backwards to build an experience that addresses that need.
• Rethink your organization structure by having a dedicated team responsible for researching and experimenting new technologies.
• Work with AWS Partners to bring the best ideas to production.

If your organization is yet to drive value from AI, consider these next steps:

• Review the AWS AI/ML Solutions page.
• Explore Amazon Bedrock and Amazon’s approach to responsible AI.
• Learn more about MONY Group’s Data Strategy for AI in financial services.
• Contact your AWS Account Manager to discuss AI partnership options. Bring insights on your biggest challenges, ambitions, and timelines on moving AI to production and start your move to AI solutions.

Firemind – AWS Partner Spotlight

Firemind is an AWS Premier Tier Consulting Partner that provides end-to-end solutions that help businesses harness data, enhance decision-making, and drive measurable outcomes. With deep expertise in Machine Learning, AI, and AWS Cloud Engineering, they accelerate AI adoption through solutions like Pulse – generative AI accelerator – and Arc for AIOps – AI workload management tool.

Contact Firemind | Partner Overview | AWS Marketplace