For Companies, AI Fluency Isn't About the Tools—It's About Judgment

As companies make AI fluency into a hiring and promotion requirement, experts warn it's less about tool mastery and more about judgment.

Paromita Gupta 8 hours ago

Topics

Key Takeaways

AI fluency is becoming a core strategy. Companies like Shopify, Accenture, KPMG, and Meta have integrated it into their hiring, promotion, and performance-review criteria.

Experts frame real fluency as knowing where AI is reliable versus risky and treating outputs as drafts, not answers.

AI fluency isn’t a level playing field: access to paid tools, fast internet, and free time to experiment shape who gets fluent.

In 2025, Shopify took a bold step by fully embracing AI. CEO Tobi Lütke announced in a memo that no new hires would occur unless employees could demonstrate that AI was incapable of performing the necessary tasks. “What we have learned so far is that using AI well is a skill that needs to be carefully learned by using it a lot. It’s just too unlike everything else,” he wrote.

He declared that the effective use of AI was now “a fundamental expectation of everyone at Shopify,” urging employees to experiment with the technology, acquire new skills, and share discoveries to make AI second nature across the company.

AI usage was now critical for performance evaluations and peer reviews to track employees’ progress. “Learning to use AI well is an unobvious skill,” he stated. He noted that employees tend to give up after writing a prompt and not getting the ideal result on the first try.

This wasn’t just a move to embed AI skills within the organization; it signaled a bigger shift: AI was becoming a core job requirement. A similar strategy has been adopted by professional services firm Accenture. “AI at Accenture is how we do work…So if you want to get promoted, you’ve got to do the things that we do to operate at Accenture,” said CEO Julie Sweet.

AI Fluency is Conceptual in Nature

For Dr. Mahmoud Mousa, Assistant Professor, School of Mathematical and Computer Sciences, Heriot-Watt University Dubai, AI fluency, the ability to interact with artificial intelligence systems effectively, efficiently, ethically, and safely, is primarily a conceptual framework, not a checklist of tool-specific skills. “True fluency lies in understanding how AI systems work, where they add value, and where their limitations lie,” he said.

A fluent person grasps the necessity of iterative refinement by treating the first output as a draft, not an answer.

Dr Mahmoud Mousa, Assistant Professor, School of Mathematical and Computer Sciences, Heriot-Watt University Dubai

True AI fluency requires more than just technical proficiency; it calls for a fundamental shift in mindset. One that acknowledges and accepts that AIs aren’t autonomous problem solvers but rather means to achieve a solution. The formula should be: AI scale + human judgment.

For example, Deloitte had to partially refund $440,000 to the Australian government after a report on the “Future Made in Australia” initiative was found to have several significant errors. These included academic citations for people who did not exist and a fake quote from a Federal Court judgment.

While the professional service firm insisted that the use of AI had no impact on the “substantive content, findings or recommendations” of the report, it published a revised version on the Department of Employment and Workplace Relations website.

“A fluent person grasps the necessity of iterative refinement by treating the first output as a draft, not an answer,” notes Dr. Mousa.

Notably, there has been a significant surge in global companies partnering with AI firms to gain access to patent models for business growth, including Deloitte’s deal with Anthropic to give its nearly 500,000 global employees access to the Claude chatbot.

“Tools such as ChatGPT, Claude, Copilot, or domain-specific agents are important, but they are only the current interface layer. The more durable capability is conceptual: understanding where systems are reliable, where they introduce risk, and where human oversight remains essential,” said Prof. Elizabeth Churchill, Department Chair and Professor of Human-Computer Interaction, MBZUAI.

Leaders must note that learning to use AI is a gradual, flexible process, not a linear path. If they treat AI fluency as just a checkbox or rely on one-time training, they may end up with employees who have access to tools but no real understanding.

Research Context

PwC AI Jobs Barometer: Over four years, the most AI-exposed companies tripled their lead in workforce productivity growth compared to the least AI-exposed companies.
Deloitte Australia: The firm partially refunded $440,000 to the Australian government after an AI-assisted report on the “Future Made in Australia” initiative was found to contain fabricated academic citations and a fake quote attributed to a Federal Court judgment.
EY-Parthenon: AI could lift economy-wide labor productivity by 1.5% to 3% over the next decade, with the largest contributions coming from tech, finance, consulting, legal, and accounting.

The Unequal Weight of AI

When employers require AI knowledge for hiring or advancement, the burden doesn’t fall equally. AI fluency depends on one’s ability to have regular, high-quality exposure to advanced tools, such as paid subscriptions like Claude Pro or ChatGPT 5.5, a reliable internet connection, and the discretionary time to experiment and fail.

For Prof. Churchill, AI fluency can be a disadvantage if mistaken purely for technical expertise. “That risks disadvantageous employees who may not have had equal access to emerging tools, experimentation opportunities, or structured training. The fairer approach is to treat AI fluency as a teachable, scaffolded capability.”

“Managers often judge fluency and may unconsciously favor certain communication styles, such as verbose prompting, technical jargon, or confidence in outputs, which can create assessment bias in this promotion criterion. Such practices can disadvantage neurodivergent employees who have different styles,” says Dr. Mousa.

However, the experts agree that assessing AI fluency should not be completely avoided. “AI fluency should not be treated as a prerequisite skill that employees are expected to acquire on their own. Instead, it should be supported through structured training, mentorship, and role-specific pathways. For example, a marketing professional might need to understand AI-driven content optimization, while a finance professional may focus on forecasting models,” he adds.

The AI proficiency skill is no longer limited to technical roles, but has become a cross-functional baseline. “Human resources departments use AI-driven insights to plan their workforce, while marketing teams analyze performance data and operational teams use AI to distribute resources more efficiently. The ability to work with intelligent dashboards and automate routine tasks while using insights to operate different functions proves AI proficiency as an essential skill that extends beyond engineering work,” says Amitabh Roy, founder, Teamtrace.

Leaders should focus on helping their teams develop confidence and judgment about when and how to use these tools, rather than simply teaching people how to use them. In light of emerging disparities, organizations should return to their drawing boards to reassess the design of the training and evaluation framework itself, not simply as an individual capability issue.

Performative Checkbox or Meaningful Skill Signal?

With corporations expected to double their AI spending in 2026—from 0.8% to roughly 1.7% of revenues—the pressure to demonstrate returns is immense. However, as the Deloitte case illustrates, more investment need not guarantee better outcomes. The real return on AI spending will not come from broader access to tools, but from building the conceptual understanding, human oversight, and iterative cultures that make those tools work reliably.

AI proficiency is no longer limited to technical roles; it’s becoming a cross-functional baseline.

Amitabh Roy, Founder at Teamtrace

All of this raises a question: Does making AI fluency a promotion requirement risk turning it into a checkbox exercise rather than a genuine sign of skill? “The risk would emerge if organizations reduce AI fluency to usage metrics,” adds Prof. Churchill.

Uncertainty lingers about whether AI fluency will become a performative exercise, particularly if organizations reduce it to certifications or superficial indicators, as tracking the number of AI tools used or prompts generated does not necessarily reflect value creation. “To avoid these pitfalls, organizations must anchor AI fluency in measurable business outcomes. This includes improvements in productivity, quality, decision accuracy, or customer experience. Leaders should also encourage reflective use of AI, where employees are expected to explain how AI contributed to their work and what limitations they encountered,” adds Dr. Mousa.

For Roy, “AI proficiency” is about applying AI to improve how work gets done. The company has been leveraging AI to interpret real-time data, automate routine workflows, and derive actionable insights from workforce metrics like productivity, time utilization, and task patterns. “Candidates should be comfortable working with AI-driven dashboards, predictive insights, and automation features to make faster, data-backed decisions,” he says.

AI fluency becomes the difference between meaningful acceleration and overreliance.

Prof. Elizabeth Churchill, Department Chair and Professor of Human-Computer Interaction, MBZUAI

Creating a Value of Productivity?

In four years, the most AI-exposed companies have tripled their lead in workforce productivity growth over the least AI-exposed companies, according to a PwC report.

However, the effect is not simple. Dr. Mousa points out that while AI fluency can be a genuine driver of productivity, its impact depends on the task and context. The effect isn’t uniform across tasks and contexts, nor is it guaranteed. “Fluency should ultimately be judged more by actual contributions to performance than by claimed familiarity with technology.”

Prof. Churchill concurs with the sentiments, calling the effect “conditional.” “Without fluency, speed can come at the expense of accuracy, accountability, depth, and quality. People may produce work more quickly, but also introduce errors, amplify bias, or create additional review burdens for colleagues and organizations.”

“In many ways, AI fluency becomes the difference between meaningful acceleration and overreliance,” she adds.

The Right AI Fluency Criterion

The requirement cannot be completely avoided, but it needs a proper checklist. For Liisa Pursiheimo, a Finland-based HR strategist, AI fluency as a promotion criterion depends on the hiring organization’s AI maturity. “If they are ahead of the curve, the expectation for AI fluency is higher. Companies that are in earlier stages of exploring and experimenting most likely don’t have an overarching AI strategy, thus have not reached the state where it would translate into competency requirements.”

In May, KPMG launched an internal dashboard to monitor employee AI usage, setting a 75% adoption target across most of its staff. Meanwhile, Meta announced that its employees’ performance will be assessed by AI-driven impact starting in 2026. Other key firms on the AI-implementation bandwagon include Cisco, Microsoft, Amazon, and Google.

For such companies, Prof. Churchill advocates sequencing AI fluency responsibly: invest in structured, role-relevant training before evaluating performance; ensure managers understand what meaningful AI use looks like, rather than rewarding only visible AI use; and monitor whether adoption and advancement rates diverge across demographic groups.

The definition should sit across three levels: literacy (understanding what AI systems are and are not), applied competence (the ability to use AI tools critically and effectively within one’s domain), and contextual judgment (knowing when AI assistance adds value and when it introduces risk). “The weighting across these levels will legitimately differ by role. A finance analyst and a product designer need different fluency profiles, but both need judgment,” she adds.

Many organizations are still working out what genuine, context-sensitive AI competence looks like. A lawyer, a designer, a software engineer, and a public-sector leader won’t need the same specific skills. Still, each one needs to grasp where AI falls short, judge its outputs with a critical eye, manage the risks involved, and apply these tools responsibly within their own fields.

An effective strategy for leaders should combine conceptual understanding, applied practice, and judgment. Meanwhile, employees should be made aware of what AI systems actually do at a functional level, be able to improve real workflows with them demonstrably, and know how to recognize problems such as bias, hallucination, privacy risk, uncertainty, or misuse.

AI Literacy at Work: The Big Picture

What started as a “nice to have” has now become a performance metric, playing a critical role in the future of organizations and how work gets done. Roy notes that AI proficiency will become as fundamental as computer literacy within the next two to three years. “As work increasingly revolves around real-time data, automation, and predictive insights, employees across functions will be expected to interact with intelligent systems daily,” he says.

According to EY-Parthenon, AI could lift economy-wide labor productivity by 1.5% to 3% over the next decade, with the largest contributions coming from tech, finance, consulting, legal, and accounting.

As technology advances and aims to make work more efficient, leaders shouldn’t lose sight of human judgment. Tools can speed things up and handle routine tasks, but they can’t think for us. No matter how advanced technology becomes, it’s the people using it — those who ask the right questions and catch mistakes — who make sure it’s actually helpful.

“The tools themselves will continue to evolve rapidly. The lasting advantage will come from people who can critically evaluate, adapt, and responsibly apply AI within their domain,” says Prof. Churchill.

What Leaders in Each Role Must Do Differently

C-Suite	Leaders should tie AI fluency to real results—productivity, quality, decision accuracy, customer experience—not to usage numbers like tool counts or logged prompts.
HR & People Leaders	Treat fluency as a teachable, supportable skill, with structured training and mentorship, rather than something employees are expected to pick up on their own. Define it across three levels — literacy, applied competence, and contextual judgment — and weigh each differently depending on the role.
Managers	Judge fluency by actual contributions to performance, not by communication style, confidence, or verbose prompting — those proxies can systematically disadvantage neurodivergent employees.
Boards & Governance	Monitor whether adoption and advancement rates diverge across demographic groups as AI fluency becomes a promotion criterion.

Topics

About the Author

Tags:

Topics

Share