Most job postings now mention AI. Very few say what they mean by it. One posting wants a controller who can use a chat assistant to speed up month-end commentary. Another wants an engineer who can put a forecasting model into production. Both say "AI experience required."
That vagueness costs you. Strong candidates cannot tell if the job fits them, and weaker ones talk their way through interviews. Here is a simple way to sort the need before you write the posting.
Start with the work, not the tool
Tools change every few months. The work a role is accountable for changes much more slowly. Before you list any software, answer three questions:
- What outcomes does this role own?
- Where does AI help produce those outcomes?
- How much ownership does the person have over the AI-supported work?
The third question matters most. Using a tool, designing a process around it and being accountable for a system are three different jobs with three different pay levels.
The four levels
1. Everyday users
AI sits inside normal tasks: drafting, summarizing, first-pass analysis, cleaning up reports. This covers most office roles in finance, sales support, HR, customer service and administration.
What to look for: specific examples of how the candidate uses AI in a week, and a clear view of where they check its output. A bookkeeper who says the assistant drafts vendor emails but never touches a reconciliation without review is giving you a good answer.
2. Workflow builders
These people shape how a team works. They set up intake, triage, reporting or scheduling so that AI handles the repetitive steps. Think operations leads, business analysts, and people comfortable with low-code tools.
What to look for: a process they changed, what it was before, what it was after, and what they did when it broke. Evidence of iteration is worth more than a long tool list.
3. System builders
Data scientists, machine learning engineers and data engineers. AI is the product of the role, not an aid to it.
What to look for: models or pipelines that people actually use in production, and the ability to explain a technical choice in terms of a business result. A strong portfolio that never left a test environment is a different hire.
4. Leaders of AI-supported teams
Managers who decide where AI fits, in what order, and how people adjust. They are rarely hands-on, but they need to understand the tools well enough to question them.
What to look for: a record of leading a team through a real change in process, including what they stopped doing and how they dealt with resistance.
Match the level to the near term
What you need today may not be what you need in 12 months. That is no reason to write a vague posting. It is a reason to ask a few questions up front:
- Is AI supporting the work, shaping the workflow, or producing the output?
- Which skills matter most over the next 6 to 12 months?
- Where will this role likely grow as AI becomes more embedded?
- Does any part of this role shrink if AI takes on more of the task?
The last question is uncomfortable, and worth asking honestly. If part of a role is likely to disappear, say so in the plan, and design the job so the person moves toward the work that stays.
It is not always a single hire
Some AI work is bigger than one person. Governance, change management and implementation can call for a project team, a contractor or a short-term specialist rather than a permanent employee. Contract talent is often the right answer when you need a defined build, a data clean-up or a pilot, and you do not yet know what the permanent role looks like.
A useful test: if you can describe the deliverable and the end date, consider contract. If the work will keep growing and needs institutional knowledge, hire permanent.
Build in room to change
Four habits keep a team adaptable without constant rewrites:
- Define roles by outcomes, not platforms.
- Hire for a pattern of learning. Ask what the candidate picked up in the last six months and how.
- Build teams with complementary strengths rather than identical profiles.
- Review each role once or twice a year and adjust the expectations.
Interview questions that separate fluency from name-dropping
- Walk me through a task you do differently now than two years ago because of AI.
- When has an AI tool given you a wrong answer, and how did you catch it?
- What would you not hand to AI in this role, and why?
Candidates who have really used these tools answer quickly and with detail. Those who have only read about them tend to stay general.
What to do next
Take your next open role and write one sentence that places it at one of the four levels. Then rewrite the posting around the outcomes at that level and cut any tool names that do not matter. If you are hiring across several levels at once, do this for each role separately. A Traverse recruiter can help you work out which level fits, and find candidates who have done that work in a real operation.