Sema4.ai 3.0 lets teams reuse approved queries and calculations in agent workflows

Sema4.ai 3.0 adds a way to reuse approved queries and calculations.

Sema4.ai announced version 3.0 of its enterprise agent platform on October 29. The release introduces Verified Queries and Verified Logic, which let teams reuse approved data access patterns and calculations in repeated work. It also adds shared workspace features and an inbox for routing incoming work. The platform changes concern what happens after an organization has already defined how a task should be performed. [Sema4.ai release](https://sema4.ai/newsroom/sema4ai-enterprise-ai-agent-platform-team-based-work-token-reduction/)

When an agent can call a procedure instead of drafting one

If staff calculate the same reconciliation figure every week, they can approve a procedure once and run it with the current data. Asking a model to invent the calculation for each run can introduce changes that a reviewer did not intend. Reusing approved logic makes the implementation easier to compare across runs. The calculation still depends on using the right data, inputs and version.

Sema4.ai says its Verified Queries consumed 80% fewer model tokens and delivered results four times faster in its comparison with repeated model-generated queries. For Verified Logic it reports 50% fewer tokens and 30% faster runs for repeat work. These are vendor measurements tied to the compared workflows. A team with fast existing queries or a higher share of one-off work could see a different result. [Sema4.ai reported measurements](https://sema4.ai/newsroom/sema4ai-enterprise-ai-agent-platform-team-based-work-token-reduction/)

What shared workflows need from audit records

The update describes activity records with before-and-after differences, as well as shared agents, files and workspaces. That combination raises a practical review question. If an agent changes a record, an operator needs to identify the approved routine used, the inputs it received and the person or event that started it. A workspace can share a procedure without giving every participant the same rights to run it.

The useful comparison is whether repeated cases reach the correct result more consistently, including after a procedure changes. Tracking exceptions alongside token use would show where the platform still needs human judgment.