Notifications
Clear all
Topic starter
06/06/2026 2:28 am
TL;DR: At HumanX 2026, Omni CEO Colin Zima argued that analytics agents are still lightly autonomous, best suited to well-defined tasks, and that AI accelerates but does not replace the need for stable UI and human curation, according to WorkOS. The practical lesson is that governance must track where automation ends and decision authority begins, especially as software blends natural language, scheduled actions, and direct manipulation.
NHIMG editorial — based on content published by WorkOS: Modern analytics in the age of agents
Questions worth separating out
Q: How should security teams govern AI agents in analytics workflows?
A: Security teams should govern analytics agents by separating scheduled automation from genuine autonomy, then applying the right identity model to each.
Q: Why do AI-assisted analytics tools still need stable UI controls?
A: AI-assisted analytics still needs stable UI controls because users must refine outputs, compare changes over time, and understand what was approved.
Q: What breaks when analytics agents are treated as fully autonomous too early?
A: What breaks is the governance model.
Practitioner guidance
- Classify agent behaviour by runtime independence Document whether each analytics agent only follows scheduled instructions, whether it can select tools dynamically, and whether it can act without approval gates.
- Preserve stable approval and audit points Keep fixed review steps for analytics actions even when the front end is generated or conversational.
- Separate creation from curation in operating policy Write policy so that generation, validation, and deployment are distinct stages with different owners.
What's in the full article
WorkOS's full article covers the conversational and product-context detail this post intentionally leaves for the source:
- The full interview discussion on Omni's three-layer software model and how analytics interfaces are likely to evolve.
- Colin Zima's exact framing of when agents are useful, when human-in-the-loop correction is still required, and why task specificity matters.
- The examples behind the team's AI-generated code and rapid prototype culture, including how experimentation changes when the cost of building drops.
- The on-stage context from HumanX 2026 in San Francisco, which gives the conversation its original setting and audience.
👉 Read WorkOS's interview on AI agents in analytics and interface design →
AI agents in analytics: what changes for governance teams?
Explore further
06/06/2026 4:09 am
Analytics agents are not automatically autonomous just because they are AI-enabled. The article describes mostly scheduled, instruction-led behaviour with heavy human review, which fits NHI-style governance more than autonomous identity governance. That matters because teams often over-read the label and under-read the actual control model. Practitioners should classify these systems by runtime behaviour, not by the presence of an agent label.
A few things that frame the scale:
- 92% agree governing AI agents is critical to enterprise security, yet only 44% have implemented any policies to do so, according to AI Agents: The New Attack Surface report.
- Only 52% of companies can track and audit the data their AI agents access, leaving 48% with a complete blind spot for compliance and breach investigation.
A question worth separating out:
Q: How do you know if AI-generated analytics actions are operating within their intended boundary?
A: You know they are operating within boundary when every action can be tied to a known trigger, a scoped data source, and a recorded human or policy decision. If the action path cannot be reconstructed after the fact, the system is outside the control model and needs tighter governance.
👉 Read our full editorial: AI agents are reshaping analytics workflows and UI design
