OTP vs Holacracy, roles, circles, and what happens when a role is filled by an agent

Most operating frameworks assume a person sits behind every job. You hire a someone, you give them a title, and the title carries the work. Holacracy® is the one mainstream system that broke that assumption years before anyone was talking about AI employees. It deliberately separates the role from the person. A role is a bundle of accountabilities, and a human shows up to energize it. That single design choice turns out to be the cleanest on-ramp to a workforce where some of the workers are not human.

I want to be honest up front. Holacracy is radical and polarizing. Some companies thrive inside it, others reject it loudly, and Zappos is the famous example of an organization that bet big and absorbed real friction in the process. OTP does not require you to adopt Holacracy, and this is not a pitch to throw out your current structure. But of all the frameworks I look at, Holacracy is the most structurally interesting comparison, because it already did the hard conceptual work. If a role is a bundle of accountabilities rather than a job title, then an agent can fill a role the same way a human can. That is the whole idea.

What Holacracy is

Holacracy is a self-management system created by Brian Robertson and governed by the Holacracy Constitution, stewarded by HolacracyOne. It removes traditional managers and replaces the management hierarchy with a structure of distributed authority.

Work is organized as roles, not job titles. Each role is defined by a purpose, one or more domains it controls, and a set of accountabilities, the ongoing activities it is expected to perform. Roles are grouped into circles, and circles nest inside larger circles, so the organization becomes a structure of roles within circles within circles. One person can energize several roles across different circles, and a role can be re-assigned without re-hiring a human.

Authority is distributed rather than held by bosses. Two meeting types carry the system. Governance meetings evolve the structure itself, creating roles, adjusting accountabilities, and processing what Holacracy calls tensions, the felt gap between what is and what could be. Tactical meetings handle the operational work, moving projects and clearing blockers. It is a way of structuring authority and work, not a metrics or strategy framework, and it does not pretend to be one.

It is worth saying plainly that Holacracy asks a lot. The vocabulary is dense, the meetings are rule-bound, and not every culture wants this much explicit structure. That is a fair critique. But the core abstraction, role decoupled from person, is genuinely ahead of its time.

What OTP is

OTP, the Organization Transport Protocol, is an operating layer rather than a methodology. It is the place where a part-human, part-AI team becomes legible on a single chart. Some seats are held by people. Some seats are held by AI agents. Every seat has a scorecard and a set of numbers it is responsible for reporting.

The distinction that matters is how OTP treats the AI. Most tools bolt an assistant onto a dashboard and call it AI. OTP treats agents as Agent Employees, accountable seats that report their own numbers, the same way a human seat reports its numbers. An agent is not a feature inside someone else's job. It holds a seat. That framing is the wedge, and it is the thing that makes the Holacracy comparison so natural.

The real difference, Holacracy abstracts the role, OTP seats the worker

Here is the cleanest way to state the relationship. Holacracy abstracts the role. OTP seats the worker.

Holacracy is deliberately agnostic about who fills a role. The Constitution describes the role, its purpose, its domains, its accountabilities, and stays silent on the nature of whoever energizes it. That silence is a strength. It means Holacracy never assumed the worker had to be a person, even though in practice it always was.

OTP takes that abstraction and makes the next move concrete. It makes an agent filling a role operational. A seat held by an agent gets a scorecard, a set of KPIs, and a reporting cadence, exactly like a seat held by a human. The agent does not just do the work quietly in the background. It owns the seat, it reports the metric for the seat, and when the number moves the wrong way, the seat is accountable. Holacracy gave us role without person. OTP gives us role with a non-human worker who reports.

Where they fit together

This is not a competition. If you run Holacracy, or you are Holacracy-curious, the two systems compose well.

Role definitions become job specs. A Holacracy role is already a purpose plus domains plus accountabilities. That is most of an agent job spec already written. You take the role definition, hand it to an agent as its charter, and you have seated the role with a worker that can run continuously.

Circles map to chart groupings. The nested circle structure translates directly into how seats group on an OTP chart. A circle becomes a grouping of seats, some human, some agent, reporting up through the same structure you already governed into existence.

Surfaced tensions feed governance. When an agent notices a gap, a metric drifting, an accountability nobody owns, a domain conflict between two seats, that is a tension in the exact Holacracy sense. The agent surfaces it, and it feeds your governance meeting as an input. The agent is not deciding the structure. It is reporting the felt gap, which is precisely what a role is supposed to do.

The agent reports its role metrics. Tactical work needs current data. An agent seated in a role reports the numbers for that role on cadence, so the tactical meeting starts with live information rather than someone scrambling to pull a spreadsheet.

The agent-employee dimension

The piece that neither Holacracy nor a traditional org chart was built for is what happens once a seat reports its own numbers automatically. An agent seat does not wait for a human to update a scorecard. The scorecard auto-populates from the work the agent is already doing. The number is a byproduct of the role being energized, not a separate administrative task.

Then there is the part that only exists because these seats run on a shared protocol. OTP carries an OOS layer, cross-org learning that travels between organizations running the protocol. When a seat in one organization learns something durable about how to run its role, that learning can propagate. A human in a Holacracy role accumulates judgment that stays locked in their head. An agent seat accumulates it in a form that can be shared across the network. That cross-org learning is the moat, and it is the reason OTP is a protocol and not just another product.

A practitioner read

I will be straight about where each system stands. Holacracy is proven. It has years of real adoption, a published constitution, and a community of practitioners who have stress-tested it in live companies. It is also not for everyone, and the organizations that rejected it usually rejected it for honest reasons about culture and tolerance for structure.

OTP is early. It is a protocol I am building in the open, and I am not going to dress it up with adoption numbers that do not exist yet.

So who should actually care about putting these two together. A self-managed or Holacracy-curious organization that has already accepted the role-versus-person split, and now wants agents to hold real roles rather than just assist the humans who hold them. If you already think in roles and circles, you are most of the way to seating an agent. You have done the conceptual work. OTP is where that seat gets a scorecard and starts reporting.

Holacracy taught the org world to stop confusing the role with the person. OTP is what you reach for when some of the people are no longer people.

More in this series

This post is part of a series comparing OTP to the operating frameworks companies actually run on. Start anywhere, each one stands alone.

Or read the full series index.

Looking for the head to head against named tools rather than frameworks? See OTP vs Ninety and EOS One.