Designing beyond the org chart
What happens when an organization is designed around accountability— not just employees?
I explored an organizational model that connects internal employees, external expertise, intelligent agents, and the work they collectively perform—while keeping human responsibility visible throughout the system.
AI and external expertise can perform work. Accountability remains human.
The organization is more than an org chart.
Traditional organizational models are good at showing where people sit. They are much less effective at showing how responsibility, external expertise, intelligent systems, work, and resources actually move through the organization.
This structure tells us who reports to whom.
The question changes from “Where does this person sit?” to “How does this organization operate?”
Organizational structure still matters. But structure alone cannot explain work.
Start with the human organization.
Before adding contractors, intelligent agents, or project economics, the system needs a stable foundation: the people who hold organizational responsibility.
Employees form the permanent internal hierarchy.
EMP-01
Director of Engineering
Organizational authority
EMP-02
Engineering Manager
Technical leadership
EMP-05
Project Manager
Project accountability
EMP-07
Resource Manager
Capacity + allocation
EMP-03
Senior Project Engineer
EMP-04
Civil Engineer
EMP-06
CAD / Design Specialist
EMP-08
Field / Survey Engineer
Expertise can enter without entering the hierarchy.
Organizations often need specialized expertise or additional capacity without creating a permanent internal role. Contractors extend the workforce through the work itself.
Contractors connect to projects and work while an internal employee retains organizational accountability.
EMP-05
Project Manager
Holds internal accountability for the contracted work relationship.
Drainage Improvement
CON-01
Hydraulic Engineer
Hydraulic analysis
CON-02
CAD Specialist
Drawing production
CON-03
Survey Specialist
Field survey support
CON-04
Environmental Engineer
Environmental review
AI belongs to a person—not the hierarchy.
Intelligent agents can expand operational capacity, but they need a clear place in the organization. I treated each agent as an extension of a specific employee rather than as an independent organizational actor.
Every active AI agent remains directly connected to one accountable human employee.
Engineering Manager
Owns technical responsibility and reviews agent-assisted engineering work.
Checks drawing packages against defined criteria.
Supports technical research and specification lookup.
Flags inconsistencies for human review.
Project Manager
Owns project coordination and approves agent-supported planning outputs.
Supports sequencing, milestones, and schedule checks.
Synthesizes project activity into structured summaries.
Surfaces potential project issues for evaluation.
Human ownership remains explicit.
Accountability cannot terminate in another agent.
Structure becomes useful when it connects to work.
The organizational hierarchy and the work structure are different systems. The model becomes operational when the people accountable for the organization connect to the projects, deliverables, and tasks being executed.
Employees bridge organizational responsibility and project execution.
Where responsibility lives.
Project Manager
Holds organizational and project responsibility.
The employee remains inside the permanent organizational hierarchy. The agent does not.
Where execution happens.
Update drainage plan
A task can connect internal employees, external expertise, and employee-owned agents without treating them as the same kind of organizational actor.
Senior Project Engineer
Assigned to project work through the internal organization.
Contract Hydraulic Engineer
Connected through project work rather than the employee hierarchy.
Design Review Agent
Assists work under direct ownership of a human employee.
The organization exists to deliver something real.
Organizational structure becomes meaningful when it is connected to an outcome. To test the model, I placed it inside a civil engineering project where internal teams, external specialists, and intelligent agents contribute to the same body of work.
The project gives organizational relationships a shared operational context.
A real project requires more than organizational structure. It requires coordinated expertise, decisions, work, review, and delivery.
The same project can be viewed through three different layers. Each answers a different operational question.
Who is responsible?
Internal roles establish authority, ownership, and accountability for delivery.
Who or what can contribute?
Employees, contractors, and employee-owned agents provide different forms of project capacity.
What must be delivered?
Projects become deliverables, work packages, and tasks that can be assigned and evaluated.
Organizational capacity becomes engineered work.
Inside the project, abstract relationships become specific assignments: analyze conditions, develop drawings, coordinate technical decisions, update plans, and prepare work for review.
Completed work still needs a decision.
A task can be executed by an employee, contractor, or intelligent agent without automatically becoming accepted project work. Execution records what was produced. Acceptance records a human decision about whether that work is ready to move forward.
Execution can be distributed. Acceptance remains an explicit point of accountability.
The work was performed.
An assigned actor produced the requested output. The system can record who or what contributed, when it occurred, and what was produced.
The work can move forward.
An accountable human reviewed the output and made a decision: accept it, return it for revision, or prevent it from progressing.
The first output does not have to be the final output.
Technical work often moves through review and revision before it becomes accepted project work. Instead of treating this as failure, the system preserves the review cycle as part of the work history.
The system makes the review cycle visible instead of collapsing production and approval into a single “complete” state.
Output produced.
Output evaluated.
Return to active work.
Count toward accepted work.
The system records the decision. It does not make accountability disappear.
Acceptance does not claim that work is objectively perfect. It records that an authorized human reviewed the available output and made the decision to move it forward.
Every outcome consumes something.
Once work is structured and accepted, the system can connect execution to the resources required to produce it. Hours, contracted capacity, and project costs become part of the same operational picture.
Progress tells us what was accomplished. Resource consumption tells us what it took.
Task
A defined unit of project execution.
Hours
Labor or contracted capacity used by the work.
Cost
The economic consequence of consumed resources.
The project now has two stories.
One describes how much work has moved forward. The other describes the resources consumed to get there. They influence one another, but they are not the same measure.
Accepted work
How much of the approved project scope has reached an accepted state.
Budget consumed
How much of the approved project budget has been consumed by work performed so far.
40% complete does not mean 40% of the budget should be gone.
Projects rarely consume resources in a perfectly linear relationship with completion. Early investigation, mobilization, specialist work, procurement, or other front-loaded activity can legitimately change the cost curve.
Cost can enter the project through different relationships.
The system preserves those differences rather than treating every contributor as interchangeable labor.
Internal capacity
Employee time can be allocated across projects while organizational responsibility remains connected to the employee's permanent role.
External capacity
Contracted expertise can carry an authorized rate, available hours, actual hours, and resulting project cost.
Extended capacity
Agent activity remains connected to its human owner and the authorized work it supports rather than being treated as independent organizational labor.
Cost only means something in relation to a plan.
Rather than judging project performance from a single burn ratio, actual activity can be compared with the approved budget and the project's expected cost pattern.
A difference between planned and actual consumption becomes something to investigate—not an automatic judgment about project or workforce performance.
Project health is not one number.
Once structure, work, acceptance, resources, and cost are connected, the system can provide a more meaningful view of project health. But no single metric tells the whole story.
Health emerges from multiple signals interpreted together—not from a universal score.
West Oakland
Drainage Improvement
Five signals.
One project.
Each dimension answers a different question. Together they create a more useful operational picture.
What has moved forward?
Accepted work compared with the current approved project scope.
What is the work costing?
Actual and committed project costs compared with the approved economic baseline.
Where are we in time?
Time elapsed, milestone status, and schedule variance against the approved plan.
Can the project move forward?
Dependencies, approvals, blocked work, and other conditions that affect delivery readiness.
Do we have what the work requires?
Required capability compared with available and allocated project capacity.
The value comes from relationships between signals.
Individual metrics become more meaningful when they can be interpreted against the rest of the project.
Accepted work is moving forward.
Progress and elapsed time are relatively close.
Consumption is ahead of both progress and elapsed time.
Is higher consumption expected at this stage of the project?
The difference becomes a signal to investigate against the approved baseline, committed costs, remaining work, and expected cost curve. The interface surfaces the condition; it does not invent the explanation.
A project can be nearly complete and still unable to move forward.
Percentage complete describes accepted scope. It does not tell us whether a critical approval, dependency, or unresolved responsibility is preventing delivery.
Health should help us look forward.
Historical metrics explain what has happened. Project decisions also require an estimate of what remains.
Actual cost
Time elapsed
Required capacity
Open dependencies
Schedule outlook
Projected margin
A forecast is an informed projection based on the current baseline and remaining work—not a guarantee.
Intelligence
No magic health score.
Reducing progress, cost, schedule, readiness, and capacity to a single number would make the interface simpler—but the underlying decision less transparent. I chose to preserve the signals so people can understand what is driving the project's condition.
11 Architecture → Interface
The interface needed to catch up with the idea.
The original interface represented the organization. The redesign needed to represent how the organization actually operates—across accountability, work, resources, intelligent systems, and evidence.
The interaction shift
The architecture stopped being something to explain. It became something the user could navigate.
From org chart to operational organization.
Selecting a person no longer reveals only a title and reporting line. It reveals an operational footprint—projects, accountability, owned agents, external relationships, and current work.
Design consequence
Once the system could connect people to work, the interface could begin answering why.
Metrics became inspectable intelligence.
A project signal should not ask the user to trust a number. The interface exposes the source, calculation, context, and interpretation behind the value.
Organizational relationships became navigable work.
Project → Deliverable → Work Package → Task creates the execution structure, while employee, contractor, and AI relationships remain visible without pretending they are equivalent.
Activity became evidence.
The operational history preserves more than an activity feed. It records who or what acted, the related work, the state transition, and the human accountability surrounding that event.
Product principle
The interface does not hide complexity by collapsing it. It makes the relationships behind a decision inspectable.
Explore the interaction model.
The prototype explores how organizational structure, project work, evidence, resources, and human accountability can operate as one connected experience.
12 Reflection
What the model revealed.
The exploration began with organizational structure. Following the relationships through work, AI, external expertise, acceptance, and economics revealed a more fundamental design problem: how responsibility survives complexity.
Primary insight
Designing for AI did not remove the need for human accountability. It made the need to model accountability more precisely impossible to ignore.
Membership and participation are different.
Employees belong to the organizational structure. Contractors participate through work. AI agents extend the capacity of a specific employee. Treating all three as equivalent resources would erase relationships the organization needs to understand.
Doing work and accepting work are different.
Execution can be distributed across employees, contractors, and intelligent agents. Acceptance is a separate decision with its own authority, evidence, and history. Activity is not the same thing as trusted progress.
Progress and consumption are different.
A completion percentage becomes meaningful only when considered alongside cost, elapsed time, remaining capacity, dependencies, and the approved baseline. The system should surface those relationships without turning them into an automatic judgment.
Complexity can live beneath a simple interface.
The underlying architecture can remain rich without appearing on screen all at once. Progressive disclosure allows the interface to reveal the relationship needed for the decision at hand while preserving the deeper system underneath.
What survived the complexity
After expanding the model and stress-testing its edge cases, the core architecture became simpler—not because the underlying system was simple, but because its organizing principle was clear.
Organization
Human Accountability
Work
Evidence
Decision
preserve the context around every relationship.
The larger question
How can an organization expand who—or what—can perform work without losing sight of who remains responsible for it?

