What Your AI Agents Don't Know About the People They Represent

The AI agent drafts a proposal. It has the right title on it, the right vendor history, the right budget range. It goes to the right people at the right time. There is one problem: six months ago, the leader it represents privately signaled a pivot away from that vendor. That signal never made it out of a strategy session. The agent didn't know. The proposal did.

Agents Act on Behalf of People. They Just Don't Know Those People.

Enterprise AI agents are now handling real work: drafting communications, screening candidates, routing approvals, summarizing negotiations. They do this on behalf of specific people. But when you look at what data those agents actually have access to, you find org chart position, calendar context, email metadata, and past purchasing patterns. The surface, not the substance.

What's missing is the layer underneath: how this person actually communicates under pressure, what values they have committed to, which decisions they would want to review personally versus delegate without question. That layer exists. It lives in coaching notes, strategy sessions, and facilitated workshops. It just doesn't live anywhere a machine can read.

The scale of this gap is striking. 72% of enterprises have AI agents in production, yet 60% lack any formal governance structurefor those agents. Organizations are deploying judgment at scale. They just haven't decided whose judgment it is.

Three Problems. One Root Cause.

The identity data gap creates distinct failures for three different groups, but the root cause is the same in each case: identity data doesn't belong to the right people, doesn't persist, and doesn't compound.

For individuals: your employer runs a personality assessment. The results live in their HR system. You leave the company and the data stays behind. The next employer starts from scratch. Your self-knowledge doesn't travel with you. It decays where it was created.

For organizations: they deploy AI agents into hiring, communications, and approval workflows with no durable answer to the questions that matter. Who authorized this agent to act? What did it know about the people involved, and when? Can you prove the decision aligned with stated values? Only 7.2% of organizations have a named individual with formal accountability for AI agent behavior. And 35% admit they could not shut down a rogue AI agent if one emerged.

For the talent market: people optimize resumes for job descriptions. Organizations filter for credentials. Neither side is actually asking whether their identities align. The result is mismatched hires, culture drift, and burnout that could have been predicted from the data — if the data had existed in a form anyone could query.

The question isn't whether your agents know your org chart. It's whether they know your people.

The Fix Starts with a Different Principle

HR tech has spent two decades building systems that capture identity data and store it in the employer's database. That design choice is the problem. Only 26% of organizations have comprehensive AI governance policies in place, but even among those that do, the underlying data model is broken: the person is not in control of their own identity record.

The infrastructure layer that fixes all three problems starts from a different principle: identity belongs to the person. The individual holds the canonical record. The organization gets permissioned access for the duration of the relationship. When someone leaves, the record leaves with them. Every engagement deepens the same record rather than starting a new one. The data compounds instead of decaying.

This is the concept Holistic Consulting and Orion Growth are designing — called VTM HIOS, for Holistic Identity Operating System. It is not a product you can purchase today. It is a framework, a schema, and a design process we're developing with early partner organizations now. When it exists, an agent drafting a proposal would be able to query a structured, auditable identity record and act accordingly — or flag the decision for review because the record indicates this type of commitment requires personal approval. That is not a feature. That is governance. The next post in this series lays out exactly what the architecture looks like and how it gets built in phases. If your organization is deploying agents into real workflows and hasn't answered those three governance questions yet, start the conversation with Holistic Consulting.