Agents are breaking AI economics, and the fix starts in the data layer
When companies talk about the cost of artificial intelligence, the conversation almost always starts with models: which one to use, what it costs per million tokens, and whether a smaller model is “good enough.” That made sense when AI meant a chatbot. It makes much less sense now that agents are doing real work in […] The post Agents are breaking AI economics, and the fix starts in the data…
Artificial intelligence (AI) agents are changing the economics of AI by introducing unprecedented demands on the data layer, according to Couchbase Inc. senior vice president of AI, Tim Finley. Finley argues that the focus on models – such as which one to use and their per-token costs – has been misplaced, as agents perform complex tasks by querying multiple systems, invoking models several times, and writing results back.
This results in agents, rather than humans, generating 450% more traffic per task and often exceeding machine traffic on the web. Finley warns that capacity plans built for human growth may fail under the demanding workload of agents, which never sleep and perform parallel tasks. The trade-off between model choice and capabilities is a significant concern, as customers may downgrade models to reduce costs once they scale, leading to degraded results.
Couchbase's announcements, including native Amazon Bedrock integration, semantic caching, agent memory, and vector search, can help mitigate these issues. Semantic caching reduces large language model calls by 40-60%, agent memory reduces token usage by 33-45%, and vector-based context retrieval produces prompts about 70% smaller than those from full-context approaches.
By converging tool capabilities into a single copy of data, organizations can cut token waste and improve AI speed and accuracy.
Written by urgent.news from SiliconANGLE's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
