Concept · Human/AI Integration

The boundary between human
and AI is a design decision.

Henry specifies, for each atomic step in your operation, who should perform it — human, AI, or hybrid — and how the handoff between them works. The Execution Plane is where human-AI harmonization becomes concrete.

The integration problem

Why does human-AI collaboration
feel so fragile?

Most organizations adopt AI tool by tool, task by task. Someone uses ChatGPT for drafting. Someone else uses Copilot for code. A team experiments with an automation platform. The result is a patchwork — AI augmenting individual tasks without a system design behind it.

The fragility comes from undefined boundaries. Nobody has specified where AI starts and human judgment begins for each step. Nobody has defined the handoff protocol. Nobody has designed the quality gate between AI output and human review. The integration is ad hoc, and ad hoc integrations break under load.

Harmonized system design is the practice of making these boundaries explicit — at the step level. For every atomic unit of work in your operation, the Execution Plane answers: Who performs this step? In what mode? With what oversight? Under what conditions does the execution mode change?

This is the layer most organizations skip. They know what work gets done (the Work Plane). They sometimes specify how it should feel (the Experience Plane). But the Execution Plane — the layer that defines the human-AI boundary — is left implicit. Henry makes it explicit.

Without execution design

AI adoption is tool-by-tool. Boundaries are implicit. Quality depends on individual judgment about when to trust AI output and when to override it. Every human-AI interaction is an improvisation.

With Henry

Every step has a specified execution mode. The boundary between human and AI is a design decision documented in the Execution Plane. Quality gates are defined, not improvised. The system is harmonized — not patchworked.

The execution plane

What does the Execution Plane
actually specify?

Who performs the step

A named role, an AI agent, or a defined hybrid. Not "the team" or "AI." A specific executor with specific capabilities. The specification names the actor and the reason for the assignment.

How it gets done

The method, tools, and resources required. For AI execution: the model, the prompt structure, the tool access. For human execution: the procedure, the tooling, the reference materials. For hybrid: both, with the boundary defined.

When it fires

The trigger condition. Event-driven, scheduled, manual, or dependent on upstream completion. The Execution Plane doesn't just specify the executor — it specifies the activation pattern.

In what mode

Human-only, human+AI assist, AI+human oversight, or fully autonomous. This is the delegation spectrum — the core design decision for every step in a harmonized human-AI system.

The full specification

How do the three planes work
together?

The Work Plane defines what gets done. The Execution Plane defines who does it and how. The Experience Plane defines what it should feel like. All three are required for human-AI harmonization — because delegation without quality design produces efficient mediocrity.

Work Plane

Intent, inputs, outputs. What this step accomplishes, what it consumes, and what it produces. The strategic "what" and "why" of every atomic unit of work.

Execution Plane

Who performs this step, how it gets done, when it fires, and in what mode — human, AI, or hybrid. The specification that makes delegation possible.

Experience Plane

What the stakeholder should feel. The emotional and perceptual design of each interaction — the layer most operations never specify but always need.

Example · Harmonized step specification

"Step: Send 90-day renewal reminder. Work Plane: Generate personalized renewal proposal based on account usage data, referencing their original goals from onboarding. Execution Plane: AI drafts the proposal (Claude), account manager reviews and personalizes the closing paragraph, sends from their own email. Mode: AI+human oversight. Experience Plane: The customer should feel recognized and valued — not processed. The reminder should reference specific outcomes they've achieved, not generic product benefits."

The assessment

How do you know the right
execution mode?

The Fit Assessment evaluates each step against delegation readiness criteria. It doesn't assume AI is better. It doesn't assume humans are necessary. It evaluates the specification completeness and the nature of the work to recommend the optimal execution mode.

Steps with fully defined inputs, measurable outputs, and well-specified quality criteria are strong candidates for AI execution. Steps requiring contextual judgment, creative synthesis, or relationship navigation stay human — and the specification says why.

The assessment is not static. As specifications improve, as AI capabilities evolve, and as your operation matures, the Fit Assessment can be re-run. Steps that were human-only last quarter might be AI-ready this quarter. The specification is the measure — not the assumption.

Specification-driven assessment

The Fit Assessment doesn't ask "can AI do this?" It asks "is this step specified completely enough for AI to do this reliably?" The specification is the gate — not the technology.

Progressive delegation

Harmonized design is progressive. Start human-only. Specify the step. Add AI assist. Improve the specification. Move to AI+oversight. Prove reliability. Remove oversight. The progression is deliberate — driven by specification completeness.

Design the boundary.
Then harmonize human and AI.

Start with a diagnosis. End with specifications that define who does what — and why.

Talk to Henry →

The boundary between humanand AI is a design decision.

Why does human-AI collaborationfeel so fragile?