What has been most effective for reducing AI token costs?

Sagnik Nandy

Chief Technology Officer · Docusign

Reduce context size

Sagnik Nandy says his team reduced the amount of context they load, which cut the number of tokens needed by almost 50%, giving them more headroom to try more things.

The answer starts at 36:55.

AM

Andrea Malagodi

Chief Information Officer · Sonar

Vortex context extraction

Andrea Malagodi cites vortex, an aspect of their solution that he says gives the agent the ability to pull context from the codebase, such as guidelines or the code graph, to reduce the LLM burning tokens by rereading code.

The answer starts at 13:25.

Dr. Justin Ubert

Division Chief (Component CISO) · U.S. Department Of Transportation

Structured data organization

Dr. Justin Ubert says organizing data in a structured, well-defined way helps reduce AI token costs over time, because it lets an AI agent get to the right data instead of sifting through everything.

Jacob Thomas

Manager of Information Security · Texas Children'S Hospital

Sandboxing

Jacob Thomas says it comes down to sandboxing that environment first, letting teams run their models and queries, and understanding the harness to see where execution loops occur so they get value.

Kevin Lee

Chief Technology Officer & Key Pursuits Leader · Nice

Partnering with the right organizations

Kevin Lee says partnering with the right organizations helps reduce AI token costs, as they can help harness those cost savings quickly and operationalize them.