AI for Pricing and Calculations: What Businesses Must Test
Large language models (LLMs) predict numbers, they don't calculate them. Businesses should use deterministic tools for all figures customers will act on, such as prices or taxes.
Businesses employing artificial intelligence for customer-facing financial calculations must exercise caution, as widespread AI adoption in small and medium-sized enterprises (77% in the U.S.) presents unique risks when AI dictates figures.
Large language models (LLMs) are predictive by nature, generating likely outputs based on data rather than performing precise mathematical operations. Figures that customers rely on for decisions—like price quotes, shipping estimates, tax figures, or payment plans—require a deterministic approach. These critical numbers should originate from tools designed for accuracy, not from the generative capabilities of an LLM.
The potential for error is significant. An AI might produce a plausible-sounding but incorrect calculation, leading to customer distrust. A notable incident involved a BMW dealership where an AI chatbot provided an incorrect vehicle buy-back price, which was actually the customer's outstanding loan balance. Such errors can erode a company's credibility.
Experts advise that any calculation requiring a consistent, verifiable answer should be handled by a deterministic engine. The AI's role should be limited to interpreting the request and selecting the appropriate tool for the calculation. This separation ensures the integrity of financial figures and safeguards against costly mistakes.