Adopting AI is the move everyone is making.
Knowing what you need, what to choose, and whether your work system stays safe,
that is the hard part.
HAOP is the framework for the hard part: anticipate and identify the hazards AI introduces, and know how to maintain control.
Examine AI deployment through three performers: the organization authors and governs the work system, the human and the AI perform inside it, and work as done emerges from all three performing together.
HAOP examines AI deployment across five connected domains. Open a card to see the question each domain brings into view.
Robotics, automated equipment, process control, machine initiation, unsafe operating envelopes, and actions that reach the physical environment.
What can the system cause to happen in the world?
In the white paper →Surveillance, scoring, pacing, work intensification, alert burden, reduced autonomy, and continuous algorithmic evaluation.
What does this system make work feel like and require from people?
In the white paper →Automation bias, skill degradation, symbolic approval, alarm fatigue, and verification demand that exceeds the conditions provided.
Can the human still perform the role the system assigns them?
The signature hazard: Verification Overrun →Closed informational loops, representation–reality mismatch: missing context, stale data, false coherence, weak anchors, and outputs that appear more complete than the evidence supports.
What keeps the representation correctable by reality?
In the white paper →Poor function allocation, procurement before purpose, unclear authority, fragmented vendor control, accountability gaps, and decisions made without the ability to intervene.
Who controls the consequential conditions of the work?
In the white paper →A person may appear in an AI-enabled workflow and still be unable to provide meaningful verification. The difference begins before the system is deployed.
A person is placed at a point in the operating workflow to review, approve, correct, pause, or intervene.
The human is visible in the process. That alone does not establish that they can understand, verify, or control what the AI is doing.
At design time, the organization decides that a human must perform a consequential verification function, and builds the workflow around that requirement.
The work must give the verifier what the verification needs: domain knowledge, evidence and context, time, authority, and a safe path when the check cannot be completed.
When the requirement to check exceeds the capacity to check, HAOP names that hazard Verification Overrun →
HAOP is a proposed framework. Pilot work examines usability, consistency, and whether the method identifies consequential conditions that existing assessments miss. That is exactly why participation matters now.
The HAOP Workbook. Operationalize HAOP with a work-design sequence and tools.
The work-design sequence the HAOP Workbook will operationalize.
The three-performer model, five hazard domains for AI-enabled work, eight operating principles, and Accountability by Control: responsibility mapped to whoever holds control over a function, its design, and its operating conditions, before failure rather than after.
One address, read by the author. Run the method against a real workflow, bring field evidence, or challenge the framework. Disagreement is a contribution.