The U.S. Department of Labor maintains O*NET — a database of 900+ American occupations — with detailed tasks, skills, knowledge, and work styles for every role from Data Scientist to Information Security Analyst. As of May 2026, there is no entry for AI Operator. So I built the profile in O*NET’s own structure, submitted it through the Occupational Code Assignment process, and will be publishing the result here. If you’re hiring this role, developing your existing people into it, or becoming one yourself, this is what the operational reality of the work is beginning to look like.

The Gap in O*NET (as of May 2026).

O*NET classifies roles under the federal Standard Occupational Classification system. Look up the data-and-analytics family at SOC 15-2051 and you’ll find three siblings: Data Scientists (15-2051.00), Business Intelligence Analysts (15-2051.01), and Clinical Data Managers (15-2051.02). Look one branch over and you’ll find Information Security Analysts (15-1212.00), Computer Systems Analysts (15-1211.00), and Computer and Information Systems Managers (11-3021.00).

None of them describes the AI Operator. The Data Scientist transforms data, interprets it, and builds the model. The BI Analyst identifies patterns, trends, and reports the data. The Information Security Analyst defends the systems. The Systems Manager runs the IT function. The AI Operator does something different from all of them: they govern an AI-augmented workflow as an accountable operating system, with chain of evidence sufficient to satisfy internal review, external audit, or regulatory inquiry.

When a role does not exist in the federal taxonomy, hiring managers improvise, salaries scatter, training programs misalign, and HR teams write job descriptions that don’t hold up in budget reviews — or in court. This article closes that gap from the field side while the federal system catches up.

How the Existing Anchors Compare.

Before reading the AI Operator profile, anchor it against the two siblings that already exist in O*NET. The contrast tells you most of what you need to know.

Dimension	Data Scientist15-2051.00	BI Analyst15-2051.01	AI Operator15-2051.03 (proposed)
Holland code	IC	CIE	ECI
Primary verb	Analyze	Report	Govern + orchestrate
Deliverable owned	Models, predictions	Dashboards, reports	Governed workflow + chain of evidence
Failure mode managed	Model accuracy	Reporting quality	Plausible-sounding AI wrongness, at scale
Risk surface	Statistical	Business decisions	Regulatory + ethical + reputational
Cross-functional load	Moderate	Moderate–High	High (Security, Legal, Compliance, IT)
People leadership	Optional	Optional	Core
Documentation as deliverable	Secondary	Secondary	Primary
Median wage (2024)	$112,590	$112,590*	$120K–$145K projected

*BI Analyst median wage data is reported jointly with Data Scientists under SOC 15-2051. Source: O*NET OnLine and Bureau of Labor Statistics, 2024. See Holland codes below.

The Critical Signal: A Holland Code Shift.

O*NET assigns every occupation a Holland interest code — a psychometric instrument that signals what kind of person finds the work motivating. The Data Scientist is IC: Investigative (study, analyze, model) plus Conventional (follow procedures, organize data). The BI Analyst is CIE: Conventional and Investigative, with a touch of Enterprising for stakeholder work.

The AI Operator’s code, in this profile, is ECI: Enterprising first, then Conventional, then Investigative. The leading letter has shifted from I to E. That is the formal signal that the AI Operator is not a deeper analyst. They are a leader-operator whose work is to direct people and systems toward a strategic outcome, with governance discipline as their operational backbone, and analytical work as the third leg — not the first.

The Day-One Job Description: Tasks 1–16.

These tasks are structured in parallel with the 16 tasks O*NET maintains for Data Scientists. They are written in O*NET’s action-verb format and ordered by frequency and importance for the role.

1. Define scope and access boundaries for AI agents and AI-augmented workflows, specifying which systems, datasets, and actions are in or out of bounds.
2. Design, document, and maintain context packages — project rules, prior decisions, organizational standards, constraints — that AI systems load on every operational cycle.
3. Establish risk-tiered approval architectures specifying which AI-produced outputs may be auto-applied, which require single-reviewer approval, and which require dual review.
4. Review, validate, approve, modify, or reject AI-proposed changes, decisions, or work products against organizational standards.
5. Maintain version-controlled records of AI-assisted actions, including model used, context provided, reviewer, and acceptance or rejection rationale.
6. Orchestrate end-to-end loops in which AI proposes, humans review, execution environments apply changes, version control records outcomes, and downstream systems synchronize.
7. Monitor AI outputs for quality regressions, scope drift, and systemic errors; recommend corrective action via context updates, model changes, or workflow redesign.
8. Develop and maintain audit-ready documentation demonstrating chain of evidence for AI-produced work.
9. Translate strategic and operational objectives into AI-workable directives and quality criteria.
10. Coordinate with subject matter experts, IT, security, compliance, and legal stakeholders to align AI workflows with policy and risk appetite.
11. Evaluate AI tools, models, and platforms against operational requirements and governance criteria; recommend adoption, retirement, or replacement.
12. Conduct post-deployment reviews of AI workflows to assess accuracy, efficiency, alignment with intent, and downstream impact.
13. Train, mentor, and coach professionals in operational orchestration practices: scope, context, approval, memory, and strategic alignment disciplines.
14. Identify problems unsuitable for AI augmentation due to ambiguity, ethical risk, regulatory constraint, or insufficient context quality.
15. Maintain standard operating procedures, runbooks (response and troubleshooting guides and incident instructions), and playbooks (strategic response, decision and scenario-based guides) for AI-augmented work as model capabilities and organizational needs evolve.
16. Recommend organizational changes — role redesign, capability development, policy updates — required to capture full value from AI augmentation.

The Five Disciplines That Define Competence.

Tasks describe what the operator does. The five disciplines describe how an operator gets good. They were laid out in the prior piece in this series; here they are again as the competence frame underneath the task list.

The Tool Stack.

The AI Operator works at a higher altitude than the model itself — but operates a real toolkit. Here is the working stack, organized in the format O*NET uses for the Tools and Technology section.

What This Costs to Hire — and Why It’s a Bargain.

The Data Scientist median wage in 2024 was $112,590 annually. The Computer & Information Systems Manager median was approximately $170,000. The AI Operator sits in between, in a hybrid technical-leadership zone, with a projected median in the range of $120,000–$145,000.

The outlook is firmly in Bright Outlook territory — much faster than average growth — because every organization that has bought AI in the last two years has discovered they need someone to govern it. There is a near-term demand spike before training infrastructure catches up.

The bargain math: a single AI Operator who prevents one consequential failure of AI-produced work — a regulatory finding, a public-facing error, a defensible audit gap — covers a year of compensation in a single avoided incident. Most organizations are unconsciously paying this cost in risk. They should consciously pay it in salary instead.

The Submission to O*NET.

The U.S. Department of Labor accepts public submissions for occupational code review through the Occupational Code Assignment (OCA) process at onetcenter.org/coding.html. The OCA Form Part A collects industry, title, tasks, work activities, knowledge areas, tools, education, and training for the proposed occupation. After submission, an analyst at the National Center for O*NET Development reviews the request and responds within 14 business days.

I have prepared and submitted a complete OCA Form Part A package for the AI Operator role using the profile above. Here’s what I submitted Proposed Occupation: AI Operator A single submission does not produce a new O*NET-SOC code on its own — the federal system requires evidence of frequency and prevalence across multiple sources before a new code is created. But the submission does three concrete things: it adds AI Operator to the O*NET Alternate Titles file, it feeds the next occupational classification review, and it puts a documented field-level definition into the federal record. If your organization employs AI Operators (under whatever title), trains them, or hires them — consider making your own submission. Every additional voice in that review shortens the gap between what the work actually is and what the federal occupational record says it is. My work is CC BY-NC 4.0 and meant for public open source occupational workforce architecture