How to write an AI-enabled job description without creating a five-job role
The practical rule is simple: define one primary outcome, name the work that supports it, specify where AI or other tools may help, and state what remains outside the role. If a description needs three different managers to explain it, it is probably three jobs.
PwC’s June 15, 2026 Global AI Jobs Barometer analyzed more than one billion job advertisements across 27 countries and found that AI is changing the skills attached to roles, including stronger demand for judgment, creativity, leadership, and other human-intensive skills. That is a reason to redesign role scope, not a reason to add every emerging skill to one posting. [1]
Start with the outcome, not the tool
A weak opening says:
We need an AI-savvy operator who can automate everything, own reporting, support customers, manage vendors, and improve sales operations.
That is a wish list.
A stronger opening says:
This role owns the reliability of our customer-operations handoffs, from intake through resolution, so customers receive consistent answers and managers can see where work is stuck.
The second version gives candidates and managers something to organize around. Tools can change. The outcome remains legible.
The four boundaries every AI-enabled role needs
1. One primary outcome
Choose the result the role is accountable for. Examples:
Keep customer cases moving to the right owner.
Produce a dependable weekly operating report.
Turn qualified demand into well-prepared sales conversations.
Keep recruiting coordination organized and candidate communication timely.
A role can support several outcomes, but one should be primary. The others should be clearly secondary or shared.
2. A short list of owned work
List recurring responsibilities that directly support the outcome. Aim for five to seven meaningful responsibilities, not every task the team has ever needed.
Separate:
Own: the person makes decisions and closes the loop.
Contribute: the person supports another owner.
Use: the person operates an approved tool or workflow.
Escalate: the person identifies a risk or decision that requires another role.
This prevents “help with” from becoming invisible permanent scope.
3. Tool access and data boundaries
Name the systems the person will use, but do not confuse access with ownership. A role might use a CRM, ticketing platform, spreadsheet, knowledge base, and approved AI assistant. It does not automatically own the data model, vendor relationship, privacy decision, or final customer communication.
Write the boundary plainly:
The role may use approved AI tools for first drafts, categorization, and internal summaries. It may not enter confidential customer information into an unapproved system or send externally without the required review.
4. Non-goals
A short “this role does not own” section protects both sides. It can say:
Does not set company-wide AI policy.
Does not own revenue forecasting.
Does not replace the subject-matter owner for regulated or high-risk decisions.
Does not manage the entire customer-success function.
Does not promise automation savings before a workflow is measured.
Non-goals are not a lack of ambition. They are operating clarity.
A reusable job-description structure
Use this sequence:
Role purpose: one paragraph connecting the role to a business outcome.
What this person owns: three to five outcome-linked responsibilities.
Where tools help: approved systems, AI-assisted tasks, and expected documentation.
Where human judgment stays: decisions, approvals, sensitive data, and customer impact.
What is outside the role: non-goals and escalation boundaries.
How the team works: manager, close collaborators, communication rhythm, and decision rights.
Evidence of a strong fit: relevant experience and examples of ownership—not a generic list of every AI tool.
Practical example: customer operations coordinator
Role purpose
Own the flow of customer questions from intake to resolution, keeping cases organized, responses consistent, and unresolved issues visible to the customer-operations lead.
This person owns
Triage and routing of incoming requests.
Maintaining accurate case status and next steps.
Drafting internal summaries and response options.
Updating approved knowledge-base content when a recurring gap is confirmed.
Escalating sensitive, unusual, or unresolved cases to the designated owner.
AI-enabled work
Use the approved assistant to summarize long case histories.
Draft response options for a human to review.
Group recurring questions to help prioritize documentation.
Flag missing information before a case is assigned.
Boundaries
The coordinator does not approve refunds, change contract terms, or make policy exceptions.
The coordinator does not paste confidential customer information into unapproved tools.
Final customer-facing responses for sensitive cases remain with the customer-operations lead.
Automation proposals are documented and reviewed before deployment.
This is a real role. It uses AI, but it is not simultaneously a support manager, data engineer, compliance officer, and product manager.
Final check before publishing
Ask four questions:
Can one manager explain the role in two minutes?
Is there one primary outcome?
Are tool access and decision rights separate?
Could a reasonable candidate tell what they would not be expected to own?
If the answer is no, narrow the role before adding more qualifications.
FAQs
Should every job description mention AI?
No. Mention AI when it changes the work, the tools, the controls, or the expected outcome. Avoid adding it as a decorative requirement.
How many responsibilities should a role have?
Use a short list of outcome-linked responsibilities. If the list keeps expanding, separate primary ownership from collaboration and create non-goals.
Should candidates be required to know specific AI tools?
Only when the tool is genuinely required on day one. In most cases, describe the workflow and the judgment required rather than locking the role to a fast-changing brand name.
Who should approve the final job description?
The hiring manager should own the scope, with input from the people who will work with the role and any security, privacy, or compliance owner affected by the workflow.
A well-scoped role gives an AI-enabled team room to improve without asking one person to become the company’s entire automation department.