Join OTP the operating platform for people and AI agents
Back to Blog
Founder Notes 2026-06-16 · David Steel

What Changes When Agents Do Half the Work

When agents do half the work, the unit of management shifts from the person to the seat. Work no longer flows through a fixed headcount. It flows through a network of human and agent owners that has to be governed, measured, and held accountable as one team. The org chart stops being a record of who reports to whom and becomes the operating system for how output gets produced.

This is not a distant scenario. BCG reports that 35% of organizations have adopted agentic AI, with another 44% planning to deploy soon, and that 65% of managers expect agents to take over at least half of their job within three years. The change is arriving inside the current planning horizon, not the next one.

Accountability becomes the scarce resource

The first thing that breaks is ownership. When a human did the work, accountability traveled with the job title. When an agent does the work, the question "who owns this outcome" no longer has an automatic answer. BCG frames agents as having a dual nature, part tool and part teammate, and that duality is exactly the problem. A tool needs a user. A teammate needs a seat, a manager, and a measure. Treat an agent as only a tool and you lose the accountability that made the work trustworthy. Treat it as a fully autonomous teammate and you lose control. The executive job is to give every agent a named owner and a clear accountability, the same way you would for any human hire.

Coordination cost moves to the center

In a company where agents do half the work, the bottleneck is no longer execution. It is coordination. Agents can produce drafts, analyses, and decisions faster than the organization can route, review, and reconcile them. Output that nobody integrated is not output. So the value shifts to the connective layer: how seats hand off to each other, how priorities are set, how issues get surfaced and resolved, and how the whole system stays honest about what is actually true. Companies that win this transition will be the ones that build that coordination layer deliberately instead of letting it form by accident in chat threads and inboxes.

Management itself gets restructured

If most managers expect agents to absorb half their role, then management is being redefined in real time. The repetitive coordination, status-chasing, and first-draft work moves to agents. What remains for human managers is judgment, accountability, and the design of the system itself. That requires a maturity path, because no organization jumps from agents-as-tools to agents-as-teammates overnight. It moves through stages, from simple assistance to supervised autonomy to teams of agents that coordinate with light human oversight. Naming where you are on that path is the difference between scaling agents on purpose and accumulating them by drift.

OTP exists for exactly this moment. It puts every seat, human and agent, on one org chart with a clear owner and accountability, adds the scorecard, priorities, and issues that keep a hybrid team coordinated, and tracks progress along OTP's 8 Levels of agentic maturity through a governance layer called the OOS. When agents do half the work, you do not need another tool. You need an operating model you can run. See it at orgtp.com.

DS
David Steel

Founder of OTP. Runs an AI agent army at a digital agency. Building OTP because nobody else seems to be building it. Notes from inside the build, not from the conference circuit.

More about David →

More posts on the blog index.

All posts