Modeling AI Agents as a Class of Non-Human Identity — An Ontological Approach to Enterprise Governance

[Frank Guerino]

Hi All,

 

I’d like your thoughts, please…

 

As enterprises deploy AI agents at scale, a governance challenge emerges that I believe has a genuine ontological dimension: How do we formally classify AI agents as a distinct class of entity within the enterprise model, and how do we define their typed relationships to every other class of enterprise entity?

 

I’ve attempted to address this through two openly available, vendor-neutral artifacts:

1. A structured AI Agents Inventory with defined Attributes document — covering 30+ attribute categories and explicit typed relationships to 10 other enterprise entity types including Applications, People and Personnel, Data and Information, Regulations, Environments, Capabilities, Integrations, Vendors, Software Technologies, and AI/ML Models (all as a starting point).
2. A broader Enterprise AI Governance Best Practices document that heavily relies on such an inventory and its defined attributes as the backbone for governance.

 

What I believe makes this ontologically interesting is the classification challenge itself.

• AI agents are a category of Non-Human Identity — distinct from applications (which are passive), distinct from people (who are human), and distinct from services (which are functional abstractions).
• Because of this, governing AI agents requires a formal entity model that captures their autonomy, behavioral controls, provenance, assurance, and compliance obligations in relationship to every other governed entity class.

 

I’m curious as to whether others in the community are working on formal ontological models for AI agent classification and governance, and whether the entity relationship structure I’ve shared maps to established ontological patterns anyone here has encountered.

 

Your thoughts are very much appreciated.

 

My very best,

 

Frank

--

Frank Guerino, Principal

The International Foundation for Information Technology (IF4IT)
http://www.if4it.com
1.908.294.5191 (M)

LinkedIn: https://www.linkedin.com/in/frankguerino/

[Dan Brickley]

On Thu, 30 Jul 2026 at 17:54, Frank Guerino <frank....@if4it.com> wrote:

Hi All,

 

I’d like your thoughts, please…

 

As enterprises deploy AI agents at scale, a governance challenge emerges that I believe has a genuine ontological dimension: How do we formally classify AI agents as a distinct class of entity within the enterprise model, and how do we define their typed relationships to every other class of enterprise entity?

Timely discussion. Although we speak of "agents" as if they're as individually countable as pets, livestock, humanoid robots or humans, in practice they are much more fluid, transient, ephemeral. 

I'd focus on trying to nail down the parts that are stronger candidates for having enduring identity. For example parameter snapshots, model releases and system prompt versions. All of these are under increasing scrutiny for review and approval processes, even if AI companies naming/versioning practices have historically been pretty confusing. 

For the "agentic" aspects, rather than worrying too much about when two so-called-agents are the selfsame thing, or  versions different versions or instantiations of some platonic agent variant, ... we might do better to model aspects of the protocols and agreements under which they act in the world, access data, call APIs and so on. Name the nameable and the rest matters less. So oauth-like tokens and so on rather than the metaphorical pseudo-people that models, systems, system prompts and data access tokens are combined to represent.

cheers,

Dan

 

 

I’ve attempted to address this through two openly available, vendor-neutral artifacts:

1. A structured AI Agents Inventory with defined Attributes document — covering 30+ attribute categories and explicit typed relationships to 10 other enterprise entity types including Applications, People and Personnel, Data and Information, Regulations, Environments, Capabilities, Integrations, Vendors, Software Technologies, and AI/ML Models (all as a starting point).
2. A broader Enterprise AI Governance Best Practices document that heavily relies on such an inventory and its defined attributes as the backbone for governance.

 

What I believe makes this ontologically interesting is the classification challenge itself.

• AI agents are a category of Non-Human Identity — distinct from applications (which are passive), distinct from people (who are human), and distinct from services (which are functional abstractions).
• Because of this, governing AI agents requires a formal entity model that captures their autonomy, behavioral controls, provenance, assurance, and compliance obligations in relationship to every other governed entity class.

 

I’m curious as to whether others in the community are working on formal ontological models for AI agent classification and governance, and whether the entity relationship structure I’ve shared maps to established ontological patterns anyone here has encountered.

 

Your thoughts are very much appreciated.

 

My very best,

 

Frank

--

Frank Guerino, Principal

The International Foundation for Information Technology (IF4IT)
http://www.if4it.com
1.908.294.5191 (M)

LinkedIn: https://www.linkedin.com/in/frankguerino/

--
All contributions to this forum are covered by an open-source license.
For information about the wiki, the license, and how to subscribe or
unsubscribe to the forum, see http://ontologforum.org/info
---
You received this message because you are subscribed to the Google Groups "ontolog-forum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ontolog-foru...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/ontolog-forum/105B5E26-0115-45D8-94FD-9767B37E3AFE%40if4it.com.

[Alex Shkotin]

Hi Frank,

I was discussing with Deepseek the idea that AI agents are a step forward in the evolution of software agents, which have long existed in information systems.

Here's a brief summary [1]. The works Deepseek cites lay out the theory behind this topic.

Should we formalize it?

Alex

[1] https://chat.deepseek.com/share/bpwdqs21hypmrxt58r

**A Classification of Software Agents: From Unix Daemons to AI Agents**

In current discussions about AI agents, it is often overlooked that software agents as a class of systems have existed long before the recent boom in artificial intelligence. The foundational work by Wooldridge and Jennings (1995) introduces a crucial distinction between the "weak" and "strong" notions of agency, which perfectly clarifies this issue.

According to this framework, the **"weak" notion of an agent** requires just four properties: autonomy, reactivity, proactiveness (goal-directedness), and social ability (the capacity to interact with other agents). The authors explicitly note that even systems such as **Unix daemons** — background processes that autonomously perform tasks in pursuit of given objectives — satisfy these criteria. These are precisely the kind of software agents that have operated within enterprise information systems for decades, long before AI entered the scene.

The **"strong" notion of an agent** adds to this foundation a set of intellectual or human-like characteristics — the capacity for learning, reasoning, forming beliefs, and holding intentions. These are what we now call **AI agents**, which are merely a specific, evolutionarily more advanced subset within the broader class of software agents.

In this light, AI agents are not a revolutionary replacement for everything that came before, but rather a natural development of a long-established paradigm of autonomous software systems.

**References:**

- Wooldridge, M., & Jennings, N. R. (1995). Intelligent Agents: Theory and Practice. *The Knowledge Engineering Review*, 10(2), 115-152.

- Wooldridge, M. (2009). *An Introduction to MultiAgent Systems* (2nd ed.). John Wiley & Sons.

чт, 30 июл. 2026 г. в 19:54, Frank Guerino <frank....@if4it.com>:

--