The next time one of these roles opens, you will pull up the last job description and edit it.
Almost everyone does. A req comes back, someone adds a line about the AI tool the team now uses, and the posting goes out. Nobody stops to ask whether the rest of the document still describes the job.
Start with the simplest test: delete the AI line.
If the rest of the posting still describes a coherent role, the company may have adopted the tool without redesigning the work around it.
That conclusion is not automatic. Job descriptions are imperfect records of how work actually gets done. But they are one place redesign should become visible. If no trace of it appears, either the work has not changed or the company is hiring against an obsolete description. Both are worth knowing.
This brief gives you five questions for running that check. It takes about twenty minutes and one posting.
Adoption is common. Redesign is where the return is.
Why this is worth twenty minutes
Three findings, from surveys large enough to argue with.
Most companies are not seeing the money. PwC surveyed 4,454 CEOs across 95 countries and territories. Fifty-six percent said AI had produced neither additional revenue nor lower costs in the previous twelve months. Twelve percent reported both.
Access and experimentation are no longer the main constraint. McKinsey surveyed 1,993 people across 105 countries. Eighty-eight percent said their organizations use AI in at least one business function. Only 39 percent could attribute any enterprise-level EBIT impact to it, and most of those put the impact below 5 percent. The tools are in the building. In most companies, they have not been built deeply into the work.
The companies seeing the strongest returns are much more likely to have changed the work. In the McKinsey survey, AI high performers were nearly three times more likely to have fundamentally redesigned individual workflows. Of 31 variables tested, workflow redesign was among the strongest predictors of meaningful business impact. PwC found the same broader pattern: the companies reporting both revenue and cost gains had embedded AI much more extensively across products, demand generation, support functions, and decision-making.
Adoption is common. Redesign is rare. The gap between them is where much of the return goes missing.
A job description is not a complete operating model. But it is one of the few documents that states what the company believes the work is. You hire against it, set expectations from it, and often carry its assumptions into performance management.
If the work changed, some evidence of that change should appear.
The mechanism: directed or absorbed
An AI tool can change a role in several ways. It can return time, shorten cycle times, or make work economical that was not economical before.
That gain then goes one of two ways:
Directed decided
Someone decides what the new capacity or capability is for.
The role gains a new output, a higher standard, a shorter cycle, or more time for judgment-intensive work.
Absorbed the default
Nobody makes that decision.
The gain disappears into the existing shape of the job. The person stays busy, but the company cannot name what it received.
Absorption is the default. It is what happens when no step in the process asks what the business expects to get from the tool.
That is not a failure by the employee or the person who wrote the posting. It is an unmade management decision.
How to use the five questions
Choose a job description your company has posted since AI tools entered the work.
Read it once. Then answer each question below in one sentence.
Each question includes a strong answer and a weak answer. Neither is a grade. A weak answer tells you where the role has not caught up with the tooling, or where the company has not yet decided what it wants from the change.
That is the useful finding.
The five questions
Changed work
What responsibility, decision, or performance standard changed because of the tool?
Read the responsibilities as if the AI line were not there. Ask which parts of the role now operate differently because the tool changed what is practical, economical, or achievable.
The work does not have to be literally impossible without AI. It does need to be materially different because AI is available.
Strong. The posting describes a responsibility, decision, cadence, scope, or quality standard that would not have been realistic in the previous version of the role.
Weak. Every responsibility would have appeared two years ago. The only material addition is a requirement to use or understand AI.
Business output
What will the company get from this role that it did not get before?
The answer may be a new output: an analysis, artifact, service, or decision the company did not previously produce.
It may also be a meaningful change to an existing output: proposals delivered in one day instead of five, twice as many accounts reviewed, fewer errors, better customer response, or a lower cost to serve.
“Faster” is not a weak outcome. “Faster” without a business consequence is.
Strong. You can name what will be produced differently and why that difference matters to the business.
Weak. The posting promises that the person will be “more efficient” but leaves the value of that efficiency undefined.
Returned capacity
When AI returns capacity, where is that capacity supposed to go?
Job descriptions rarely contain a field for returned capacity. That does not make the question optional.
The gain can be directed toward deeper review, stronger relationships, or work the role never had room for. Different roles will justify different answers. What matters is that someone has chosen one.
The capacity is worth most where it goes back into work only these people can do: the judgment, the client contact, the thinking that was always queued behind the routine.
Strong. The posting makes clear what the role will take on, improve, or stop doing because routine work now costs less time.
Weak. The posting describes the same volume of the same work and makes no decision about the capacity the tool is expected to create.
Authority
Whose job is it to change how this work is done?
Using a tool inside an existing process is not the same as redesigning the process.
This is the McKinsey finding at the scale of a single role. The companies seeing returns redesigned their workflows. Redesign requires someone whose job includes doing it.
The person in the role does not need unilateral authority. But redesigning the work has to be somebody’s named accountability. Without that, employees can identify opportunities and still not capture them.
Strong. The role carries defined authority to change its own workflow within clear boundaries, or the posting names who holds that authority and how the role reaches them.
Weak. The person is expected to “drive AI adoption,” but nobody is accountable for changing the process and the approval path is unspecified.
Accountability
What measure should move if the redesign is working?
Tool use is not the outcome. Training completion is not the outcome. Prompt volume is not the outcome.
The measure might be cycle time, error rate, customer satisfaction, or cost to serve. It does not need to be perfect. It needs to connect the changed work to something the business values.
Strong. The company can name a baseline, the direction or target it expects to move, and when it will review the result.
Weak. Success means that the employee uses the tool regularly or finds it helpful.
Next step
What to do with the answers
Take the role that has changed most since AI tools entered the company.
For the first pass, remove the tool requirements and write the role from scratch. Define:
- What the business needs from the role
- What the person can now do, produce, or decide differently
- Where any returned capacity should go
- Who has authority to redesign the workflow
- Which measure should move if the redesign works
Only then add the specific tools, capabilities, validation responsibilities, and controls the person will need.
If the new responsibilities and measures come easily, the work has probably changed along with the tooling.
If they are hard to name, that is evidence too. It usually means the company adopted the capability before deciding what business value it was supposed to create.
That decision is still available.
- Changed workWhat responsibility, decision, or performance standard changed because of the tool?
- Business outputWhat will the company get from this role that it did not get before?
- Returned capacityWhen AI returns capacity, where is it supposed to go?
- AuthorityWhose job is it to change how this work is done?
- AccountabilityWhat measure should move if the redesign is working?
Sources. PwC, 29th Global CEO Survey (19 January 2026). PwC surveyed 4,454 CEOs in 95 countries and territories between 30 September and 10 November 2025. Fifty-six percent reported neither revenue nor cost benefits from AI during the previous twelve months; 12 percent reported both. CEOs reporting both were two to three times more likely to have embedded AI extensively across products and services, demand generation, and strategic decision-making.
McKinsey, The state of AI in 2025: Agents, innovation, and transformation (5 November 2025). McKinsey surveyed 1,993 participants in 105 countries between 25 June and 29 July 2025. Eighty-eight percent reported regular AI use in at least one business function; 39 percent attributed some enterprise-level EBIT impact to AI, most of those below 5 percent. AI high performers (about 6 percent of respondents) were nearly three times more likely to report fundamentally redesigning individual workflows. Workflow redesign was among the strongest contributors in a relative weights analysis of 31 variables.