Employers want AI fluency backed by practical experience

AI fluency is becoming a test of experience

Some employers would rather leave a technology position open than hire someone whose capabilities fall short, according to Ger Doyle, ManpowerGroup’s regional president for North America. Employers are scrutinizing technology résumés even as they compete for suitable candidates. For hiring leaders and applicants, fluency in artificial intelligence (AI) increasingly means showing practical judgment when using the technology.

“The market is no longer defined simply by how many people employers hire, but by how precisely they hire,” Doyle said. ManpowerGroup sells staffing and workforce services, so it benefits when employers spend on finding and developing qualified people. Doyle’s account raises a practical question for employers: which capabilities justify a higher hiring threshold, and how will workers acquire them?

Job postings combine shorter skill lists with higher experience requirements

Changes in job postings show how that threshold is moving. Since late 2024, the average number of skills requested in technology vacancies has declined from about 30 to 21, while required experience has increased from about 4.6 to 4.8 years. Dean Boerner, a data scientist at Revelio Labs, described the changes in a study published last month. Revelio sells workforce data and analysis, giving it a commercial interest in employers’ use of hiring-market research.

The two measures show different changes in what employers request. Skill lists have contracted substantially, while the smaller increase in experience requirements asks candidates to bring more time on the job. A shorter list can therefore coexist with a higher experience threshold. Applicants need to show how their previous work meets the requirement, even when fewer technologies appear in the posting.

Revelio found a particularly strong preference for experience in data organization, data research, and enterprise architecture design, the planning of how an organization’s systems and data work together. The company interprets this as employers favoring experienced staff over assigning those responsibilities to junior workers or AI systems. In that reading, prior experience helps determine who receives work requiring judgment and responsibility.

Alongside that preference, Revelio found more specific language about software development. References to general software and application development declined, while requests for Java API development and Python development increased. An API is an interface through which software systems communicate. These narrower requirements favor candidates who can connect their experience to a specific contribution and show they can begin useful work quickly with newer tools and services.

Employers are becoming more selective in a market that still has substantial demand. CompTIA’s study, released earlier this month, recorded 272,040 new information technology (IT) openings in September and 625,633 active listings, the highest total in more than three years. New openings count additions during the month; active listings cover the larger pool available. CompTIA has a commercial interest in technology-skills development, which readers should keep in view when considering its market research.

Within that active pool, CompTIA counted approximately 349,821 jobs requiring AI skills in September, describing them as more than half of active listings. The count rose by 32,327 compared with August. These figures measure how common AI requirements are in advertised jobs. Revelio’s findings address a separate question: how employers are changing the skills and experience they request.

Employer surveys offer another view of those requirements. In its fourth-quarter talent outlook, Experis, part of ManpowerGroup, reported that 34% of US respondents in a recent survey of technology and IT service employers identified AI literacy, the ability to understand and use AI effectively, as a sought-after technical skill. AI modeling and application development followed at 33%. As a technology staffing business, Experis benefits from employers spending to fill these capability gaps.

Employers are also attaching pay increases to AI capability, according to talent consulting and staffing firm Robert Half, which benefits from recruitment demand. For its 2027 salary guide, the firm surveyed more than 2,000 hiring managers, business leaders, and employed workers. Robert Half reported: “72% are increasing pay for relevant AI skills, with 42% offering more for AI-related expertise than other technical skills,” linking relevant AI capability and advanced expertise to higher compensation.

The value of that capability depends on the work it enables. In its June study, consulting firm PwC linked sought-after AI skills to organizational priorities and results; PwC also benefits when companies buy advice on AI adoption and workforce change. Candidates therefore need to explain what their experience enables them to accomplish. Employers need to define the business contribution that makes a skill worth paying for.

AI work extends into infrastructure, verification, and compliance

Those business contributions span a broad range of jobs. “We’re seeing it influence hiring across occupations ranging from data science and engineering to project management and operational roles,” Doyle said. Hiring plans therefore need to include people who apply AI in management and operations as well as those who build AI systems.

The infrastructure supporting those systems adds another set of requirements. Kye Mitchell, head of Experis US, links AI infrastructure construction to demand for semiconductor specialists, network architects, construction managers, and software developers. Newly installed data centers and server racks need people qualified to run them, Mitchell emphasizes. Employers also seek skills in networking; cloud development, or building software for remotely hosted computing services; and machine learning, which uses data to train systems to recognize patterns.

Within software development, automation is changing what engineers must do with the code they produce. Developers are moving further into data and machine learning and taking on AI-engineering duties, including reviewing AI-generated code and satisfying compliance requirements. Reviewing that output requires judgment about whether it suits its intended use. As code generation becomes automated, responsibility for checking the result becomes more important.

Compliance gives that checking a specific purpose. “Governance is a thing in which engineers will have a key role to play… AI engineers build contextual guardrails for compliance,” said Deepak Seth, senior director analyst at Gartner. Gartner’s research and advisory business benefits when organizations seek guidance on adopting and managing AI. The guardrails Seth describes are rules and controls tailored to a system’s operating environment, so engineers need to understand both the applicable requirements and how the system behaves.

Robert Half’s salary projections put different pay ranges on these specialties. The firm expects AI engineering to be among the best-paid occupations in 2027, with its stated range varying by experience. Its figures distinguish AI engineering from AI architecture and give a collective range for AI/ML analysts, machine-learning engineers, and data scientists; AI/ML means artificial intelligence and machine learning. RPA means robotic process automation, which uses software to automate repeatable tasks.

Location also influences pay, so employers need to account for it when using these ranges. The distinctions between roles matter because architecture, analysis, engineering, data science, and automation involve different responsibilities. A workforce plan needs to identify the responsibility a person will carry before choosing the job title and pay range.

Junior workers need opportunities to develop judgment

Responsibilities such as reviewing generated code and applying compliance rules help explain why experience matters to employers. Yet junior workers need assignments that let them develop that judgment. As hands-on practice and workplace learning gain importance relative to academic credentials, graduates face an additional requirement: showing they can apply their knowledge under working conditions.

Revelio describes the disadvantage directly: companies increasingly want specialized capabilities requiring substantial judgment, “which younger workers are the least equipped to offer,” the firm said. Its finding that employers favor experienced staff for data-related responsibilities points to a further difficulty. Assigning that work to established employees gives junior staff fewer opportunities to build the experience they will need for similar assignments.

PwC’s June report identifies a related problem in entry-level jobs exposed to AI. Those positions increasingly require leadership and strategic thinking, capabilities people ordinarily develop through experience. An entry-level title can therefore carry expectations shaped by earlier opportunities to make decisions. Employers need to examine how much independent judgment a new recruit must exercise and where that person will learn it.

Graduates can still distinguish themselves by showing several capabilities together. Mitchell identifies technical fundamentals, practical experience, AI fluency, and strong communication as a combination that helps fresh graduates stand out. Practical experience remains part of that profile, alongside the ability to explain and apply technical knowledge. Employers seeking that combination need a way for promising people to develop it.

Employers need to build the capability they assess

Development matters for existing employees as well as new recruits because AI changes the work people already perform. Michael Chui, senior fellow at McKinsey, connects those changing workflows to adjustments in staffing, training, and potentially incentives. “You have to have the right talent in place. You have to retrain. You might have to change the incentives. That’s all hard work,” Chui said. McKinsey advises organizations on such changes and benefits when they purchase that assistance.

Chui’s emphasis on incentives matters when employees take on responsibilities such as checking generated output and meeting compliance requirements. Training can help them develop the necessary capability, but their jobs must also give them time and support to use it. Staffing determines who carries the responsibility, while incentives influence which work receives attention. These decisions shape whether newly learned skills become everyday practice.

For junior employees, PwC recommends directing development toward earlier responsibility. “Organizations must rethink how they mentor and train junior staff, helping them step up to complex decision-making much earlier in their careers,” the firm said. The recommendation addresses entry-level roles that already demand skills usually acquired through experience. Mentoring therefore needs to prepare employees for the decisions those roles require.

Internal development also helps employers meet wider staffing needs. “[Employers] are also building talent from within, expanding talent pools, and using multiple strategies to meet their needs,” Mitchell said. Recruitment can bring in experienced specialists while development helps existing staff qualify for more demanding assignments. Each approach addresses a different part of the organization’s capability needs.

As employers develop those capabilities, Seth expects them to make greater use of formal assessment. He forecasts that, by 2027, 75% of hiring processes will include certifications and recruiting tests for workplace AI proficiency. His projection places practical AI competence among the evidence employers may routinely request from candidates.

Seth connects that forecast to earlier technical qualifications. “Anybody in the IT world knows… we were asked for certifications about Java or C++ or whatever. Now suddenly that will shift to workplace AI proficiency,” he said. Academic credentials may carry less weight while structured testing becomes more common. Employers would still require evidence of readiness, with assessment focused more directly on workplace capability.

Certifications could give candidates another way to show that readiness, while employers continue to value practical experience and judgment. Assessment establishes whether someone currently meets a requirement; mentoring and retraining help people reach it. These activities need separate provision in workforce plans because paying to test capability does not create the opportunities through which employees acquire it.

Faster skill turnover requires continuing development

Reaching a hiring standard is one stage in a longer learning process. Seth forecasts a shorter technical-skills half-life, the time over which a skill loses half its useful value: “The half-life of technical skills will shorten from 8 to 12 years to 2 to 5 years. So adaptability and learning velocity will become very, very key,” he said. Under that forecast, employers need to assess and support how quickly people can learn as their work changes.

Faster turnover also affects the choice between developing employees and replacing them. “Rehiring is expensive. Reskilling is a better option,” Seth said. His recommendation makes continuing training part of staffing decisions: when technical requirements change, employers need to decide which employees can learn the new work and provide the time and support for them to do so.

Key highlights

• Hire for demonstrated judgment: Technology job postings list fewer skills while asking for more experience. Hiring managers need assessments tied to specific responsibilities, including applying AI tools, reviewing outputs, and making decisions.
• Plan for the full AI workforce: AI investment creates staffing needs across software, infrastructure, operations, and compliance. Workforce planners need to define responsibilities and location requirements before selecting job titles and salary ranges.
• Give junior staff supported responsibility: Rising experience requirements make entry-level development harder. Managers need to pair mentoring with assignments that build practical judgment, including supervised code review and increasingly complex decisions.
• Connect assessment with development: Certifications and recruiting tests can establish current AI proficiency; mentoring and retraining help employees reach the required standard. HR and technology leaders need separate resources for assessment, learning, and time to apply new skills.
• Make reskilling a continuing investment: Gartner forecasts that technical skills’ useful half-life will shrink to two to five years. Workforce planning needs regular skills reviews and funded learning time so existing employees can qualify for changing responsibilities.