Payments company Stax is taking a multi-LLM approach to AI — leaning on a “pay as you go” strategy to ensure that it has access to the technology it needs, while considering the most cost-effective path.  

“Our strategy is optionality,” Mark Sundt, chief AI officer and former chief technology officer at Stax, told FinAi News, noting that one LLM provider may be a better fit for certain tasks. 

READ MORE: LLM war heats up: Claude vs. ChatGPT 

For example, Stax taps Anthropic’s Claude, Salesforce’s Agentforce and Google’s NotebookLM, Sundt said. 

Stax uses each platform for different applications: 

• Claude is deployed in Excel to debug spreadsheets; 

• Employees are being trained on Agentforce to do prompts in Slack; and 

• NotebookLM is being used for retrieval-augmented generation (RAG).  

RAG is a framework used to optimize the performance of AI models, scanning internal and external knowledge bases to present relevant and accurate responses, according to IBM. 

The process used to cost thousands of dollars per month on AWS — now it can be done for much less with Google, Sundt said. 

Contemplating cost

When selecting an LLM, for now, “pay as you go” has been the right fit, Sundt said. Stax is not trying to buy seat licenses at this time. 

Paying as you go and for what you use via token pricing, “I’ll do that all day long,” he said. 

The technology continues to evolve quickly, he said. “Keep your options open, because this stuff is leapfrogging.” 

Register here for the FinAi Lending Summit, set for Oct. 7-8 in Las Vegas.