Gartner says agent costs will rise fivefold by 2028 even as tokens get cheaper. Here is how to keep them in check
Falling token prices have led many enterprises to assume AI costs will fall with them. Agents are breaking that assumption.
Sean Michael Kerner
Companies are building governed semantic layers. Most say their AI agents are getting context somewhere else
Nearly twice as many companies now name live-system queries as their AI agents' main context source. OpenAI, Keewano and Snowflake supply the missing context differently.
Sean Michael Kerner
Cohere's North 2 puts AI agents on a budget and gives them a memory
Cohere is going after two common enterprise agent problems with North 2, the new version of its enterprise agent platform. It adds user quotas, rate limits, organization-wide caps and a redesigned agent harness. According to Cohere, North 2 uses memory to keep an agent's context across sessions and gives teams a shared library of knowledge and assets to draw on. It also runs in the cloud, on premises or fully air-gapped.
Sean Michael Kerner
Cohere's Model Vault now encrypts AI inference so even Cohere cannot see enterprise customers' data
Cohere plans on making the full serving stack used in Model Vault open source so that independent auditors will be able to validate that it doesn't log, export, or leak data.
Sean Michael Kerner
Every user, device and agent gets its own table in KeewanoDB, and queries never join across them
Keewano says the reason agents cannot explain why something happened is that the sequence was taken apart before they ever queried it.
Sean Michael Kerner
OpenAI's new data agent skips the one thing rivals like Databricks are racing to publish: a benchmark
Businesses used to need a data analyst just to get an answer from their own data. OpenAI says its new ChatGPT Work agent ends that.
Sean Michael Kerner
Databricks-trained AI agents match Claude and GPT-5.6 Luna's answer quality — in half the time
Instead of running every query through the same fixed number of search steps, Databricks trained a model to decide on its own — cutting wasted search time without losing accuracy.
Sean Michael Kerner
GitHub’s HydraFusion cuts AI coding costs in every benchmark. It only matches quality in one.
The vendor's benchmark table doesn't fully back its own quality claims — and it's not the only AI company selling that story.
Sean Michael Kerner
Cohere Parse 5 loses the benchmark on points. It wins on cost per page.
Cohere's Parse 5 model balances cost and structure, offering $1.50 per 1,000 pages for efficient document parsing without sacrificing essential layout.
Sean Michael Kerner
Enterprises are overpaying for simple AI queries — Snowflake's gateway now auto-routes to cut costs up to 3x
A smaller model tries the task first and calls in a bigger one only if it can't finish the job — that's how Snowflake's new routing feature decides.
Sean Michael Kerner
Enterprises with AI context layers report agent failures at more than twice the rate of those without one
A governed context layer is supposed to stop AI agents from confidently getting things wrong. New VB Pulse research shows the opposite: it's what makes the failure visible.
Sean Michael Kerner
Nvidia's Switchyard router reshuffles AI models mid-task, cutting task costs to a third in its own tests
Nvidia is betting the real fix for AI agent costs isn't a cheaper model or a smarter router alone, but owning both under one open license.
Sean Michael Kerner