TIS helps treasury teams offload time-consuming tasks using Amazon Nova

TIS serves multinational organizations in which treasury teams navigate multiple enterprise resource planning systems, diverse banking relationships, and large volumes of financial data on a daily basis. “A key challenge for our customers has always been the amount of data that they need to sift through to make decisions,” says Jonathan Paquette, chief of strategy at TIS. “AI is like a second set of hands that can make the team vastly more productive.”

To create a tool to help treasury teams speed up tasks like reporting and forecasting, TIS engaged the AWS Generative AI Innovation Center—AWS science and strategy experts with comprehensive expertise spanning the generative AI journey. Together, the teams developed the AI Personal Assistant, a tool for accelerating data retrieval and decision-making for treasury teams. This generative AI–based tool combines an intuitive chat interface and time-saving AI features like natural language queries and smart recommendations while seamlessly integrating with other TIS solutions

TIS knew that choosing the right foundation models was critical to success, directly impacting accuracy, relevance, and reliability of AI outputs. A well-matched model means that an AI system can understand domain-specific language, handle complex queries, and deliver the required performance and compliance for its intended use case. TIS opted to use Amazon Nova Foundation Models, which deliver frontier intelligence and industry-leading price performance. “The price performance of Amazon Nova Foundation Models is fantastic,” says Florian Gessner, chief architect and VP of cloud platform at TIS.

Currently, the company uses two Amazon Nova Foundation Models—Amazon Nova Micro and Amazon Nova Lite. The company uses Amazon Nova Micro to classify the content of user inputs for the AI Personal Assistant. For example, the model can recognize whether an input is casual conversation or a question about payments. If it’s the latter, Amazon Nova Micro routes the request to a text-to-SQL agent, the output of which will be executed through a query engine. 

Both the Amazon Nova Foundation Models and TIS’s knowledge bases are hosted in Amazon Bedrock—a comprehensive, secure, and flexible service for building generative AI applications and agents—helping TIS keep its solution secure and reduce time to value. “We don’t need to worry about hosting or operating models through third-party AI vendors,” says Gessner. “Using Amazon Bedrock, we can stick to our existing compliance frameworks and simply implement AI as needed—all while creating additional value for our customers.”

 

Customer feedback on TIS’s AI Personal Assistant has been highly positive. In a survey conducted by TIS, more than 65 percent of customers said the tool could save them more than 15 minutes per person each day. This adds up quickly when the tool is used across multiple departments, from treasury teams to HR. In total, 83 percent of customers said they saw strong value in the tool, with 88 percent seeing strong potential in extending its application beyond just payments.

Crucially, TIS has found that Amazon Nova Foundation Models deliver responses that are reliable and accurate. “In our business, accuracy is critical,” says Wouter de Bie, chief technology officer at TIS. “We can’t give customers inaccurate information about payments because it will lead to incorrect decisions.”

One of the most striking outcomes of collaborating alongside AWS was the speed of development. Following an AWS Generative AI Solutions LaunchPad session, the TIS team built an initial prototype for AI Personal Assistant in 48 hours. TIS had an AWS Marketplace–ready product 45 days later.

In addition to the AI Personal Assistant, TIS has also built an AI-powered tool to help its own teams accelerate request for proposal responses and reduce database maintenance workloads. “For our customers, and our teams, we see AI as a means of helping us to deliver faster and spend less time on time-consuming manual work. It’s all about time to value,” says de Bie.

Looking ahead, TIS will continue to develop solutions that help teams improve productivity. “We’ve adopted an AI-first mindset to solving customer challenges,” says Paquette. “We’re currently exploring new use cases for reporting and forecasting and for using AI to help our customers see around corners, identify errors, and build confidence in their data. Fraud detection is another huge opportunity for us.”

“It all starts with a problem that we want to solve, and we will continue to solve problems for ourselves and our customers in the most efficient way possible,” says de Bie.