Agentic AI Is Changing IT Jobs. Are You Still Assessing the Old Role?

The job changes at the handoff between the person and the agent

Most discussions about AI skills focus on tool use. Can the person write prompts? Have they used a particular platform? Do they understand automation?

Those questions may matter. They are not enough.

Tool knowledge can be trained and can become outdated quickly. The harder questions appear when responsibility moves between the person and the AI agent.

• Define the objective and the limits of the assignment.
• Give the agent the right information and access.
• Recognize when the output is incomplete, unsupported, or unsafe.
• Decide whether to accept, revise, reject, or escalate the result.
• Explain the decision to a customer, colleague, auditor, or leader.
• Remain accountable for the final outcome.

The assessment should follow the work

Those responsibilities can involve attention to detail, problem solving, rules compliance, learning agility, communication, judgment, follow-through, and the willingness to challenge an answer that looks convincing.

The correct combination depends on the job. An IT service technician, cybersecurity analyst, software developer, recruiter, and financial analyst may all work with AI agents. They should not all receive the same assessment battery.

That is the connection to ODNA Talent. The assessment follows the work instead of forcing every job into one standard profile.

A practical example: the IT service desk after AI agents

Consider an internal IT service desk.

Before agentic AI, a technician might receive a ticket, gather information, search the knowledge base, complete routine troubleshooting, reset access, document the solution, and close the request.

Now assume an AI agent handles the early steps. It categorizes the issue, retrieves system context, recommends a fix, executes an approved routine action, and drafts the ticket notes.

The technician receives fewer simple requests. The remaining work is more difficult and often carries greater consequence.

• Review the agent's reasoning before an action is accepted.
• Detect when a routine access request may indicate a security or identity problem.
• Separate a system failure from incomplete or misleading information.
• Stop an unsafe action even when the recommendation sounds confident.
• Communicate with a frustrated employee after the automated process fails.
• Coordinate with security, infrastructure, application, and vendor teams.
• Document exceptions so the process and controls can improve.

Why the old job profile can miss the new requirements

If the employer continues hiring mainly for ticket volume, memorized troubleshooting steps, and familiarity with one platform, it may miss the capabilities that now matter most.

The old job rewarded completing the routine work. The redesigned job may place more weight on verification, exception handling, communication, and accountability.

That difference should change the assessment.

How ODNA Talent turns the redesigned job into evidence

The employer should not begin by searching for an off-the-shelf AI-readiness test. It should begin with the role.

For the IT service position, the hiring team could bring the current job description into ODNA Talent and add manager input about the new workflow. The team would identify which tasks the AI agent handles, which decisions require human review, which errors carry the greatest consequence, and which capabilities must be present at entry.

ask ODNA's Assessment Architect Mode can help organize that information and recommend relevant assessment scales with a reason for each recommendation. The authorized user reviews the suggestions and removes anything that does not connect to the job.

For one service-desk role, the evidence might include measures related to attention to detail, problem solving, rules compliance, conscientiousness, interpersonal skill, or learning. A different role could require a different combination. Technical knowledge may also need a work sample, certification review, or another appropriate method outside the assessment.

After the focused assessment is completed, ODNA Talent can bring the evidence into a practical hiring workflow.

• Fit Score summarizes alignment with the criteria defined for the role.
• Expected Behaviors helps the manager understand how assessment patterns may appear at work.
• Interview Guide provides job-related follow-up questions.
• Response Notes captures the candidate's examples and explanations.
• Interview Rating System supports more consistent evaluation.
• Candidate Comparison organizes the evidence across people being considered for the same role.
• Development Planning carries relevant findings into onboarding and coaching.

No feature makes the employment decision

No score guarantees performance. Authorized people review the information alongside experience, structured interviews, work evidence, references, and other relevant information.

What ODNA Talent does is make the chain of evidence clearer. The role requirements, assessment measures, interview questions, documented responses, and development priorities can all point back to the same job.

Do not let AI fluency become another vague hiring requirement

AI fluency is quickly becoming a phrase employers add to job postings without defining it.

That creates the same problem as asking for leadership, adaptability, or good communication without explaining what the person must do.

For one job, AI fluency may mean writing effective instructions and checking output. For another, it may mean protecting data, recognizing security risk, or knowing when an automated action requires approval. In a customer-facing position, the critical responsibility may be taking over when the automated process has damaged trust.

The phrase is not the requirement. The work is the requirement.

Before adding AI fluency to a position, answer these questions.

• What work will the AI agent complete without human intervention?
• What decisions require review before action is taken?
• Which exceptions require expertise, judgment, or communication?
• What could happen if the person misses an error?
• What should the person know on day one, and what can be trained?
• Who remains accountable when the person and the agent both contribute?

The strongest interview question may begin with an imperfect AI answer

Assessment evidence should improve the interview, not replace it.

For an AI-enabled IT role, give the candidate a realistic recommendation produced by an agent. Make it mostly correct, but include an unsupported assumption, a missing control, or a weak escalation decision.

Then watch the candidate's process.

• What do they verify first?
• Which evidence do they trust?
• What additional information do they request?
• Do they recognize the consequence of the problem?
• Do they stop or escalate at the right point?
• Can they explain the decision clearly?

Connect the interview to the role

Use the same core scenario and defined rating criteria for comparable candidates. ODNA Talent's Interview Guide, Response Notes, and Interview Rating System can help the hiring team connect that conversation to the role and document the evidence.

This is more useful than asking, "Are you comfortable with AI?" Almost everyone knows the expected answer to that question. A realistic situation shows how the person thinks when the technology is incomplete or wrong.

Security and accountability belong in the role definition

AI agents can use tools, access data, and take actions. That creates different risk than a chatbot producing text. In May 2026, the National Institute of Standards and Technology reported broad agreement among respondents that AI agents present new security threats and that existing cybersecurity practices will need to be adapted.

Security is not only a system setting. It also depends on human behavior.

The employer should define what access can be delegated, which actions require approval, what information must not be shared, what evidence must be retained, and how an incident is escalated.

An assessment cannot guarantee that someone will handle every situation correctly. It can add job-related evidence about defined capabilities when it is used responsibly and combined with other information.

The final responsibility remains with the organization and its authorized decision-makers.

The business risk is hiring for work that no longer exists

Gartner found that 85% of customer service and support leaders were adding responsibilities to frontline roles as AI reduced routine contact volume. Seventy-five percent were shifting employees into different roles within the service organization. PwC has also described a move toward broader, outcome-focused roles as AI allows people to work across tasks that once belonged to narrower functions.

The numbers will vary by industry and job. The direction is clear enough to act on: employers should review the role before the next hire, not after the new employee struggles.

The cost of a weak job definition does not appear on the assessment invoice. It appears in slower ramp-up, repeated interviews, manager time, avoidable turnover, security exposure, customer frustration, and another search for the right person.

ODNA Talent cannot promise to eliminate those outcomes. It can give the employer a more disciplined way to define the role, select relevant measures, structure the interview, compare evidence, and plan development.

That is the business case.

Bring one AI-affected role

Do not begin with a company-wide AI competency model.

Bring one position where the work is already changing. An IT service role is a good place to start because the handoffs between automation and human judgment are visible.

In an ODNA Talent demonstration, we can use that role to show how the process works.

• Review the job description and the new human responsibilities.
• Use ask ODNA to organize the requirements and recommend relevant scales.
• Review and refine the assessment direction with an authorized person.
• Examine the Fit Score, Expected Behaviors, and supporting results.
• Turn those results into structured interview questions, Response Notes, and ratings.
• Carry the relevant evidence into Candidate Comparison and Development Planning.

My recommendation

You will leave with a clearer view of what the platform can and cannot do for a real position.

My recommendation is direct: if AI has changed the work, update the evidence before you make the next hiring decision.