The AI paradox: why fluency with technology does not replace human judgment in business

An opinion piece published in Fortune by Julie Teigland, Global Vice President of Alliances and Ecosystems at EY, puts forward a thesis that breaks with the usual narrative about corporate adoption of artificial intelligence: the more capable the technology becomes, the more the market rewards precisely what AI cannot do. According to the author, after two decades at a professional services firm where human judgment is the foundation of the business, that same skill — knowing how to decide, prioritize and make sense of context — is now taking hold across all kinds of roles, whether analyst, technician or machine operator.

The starting point is a figure the author flags as a symptom of the problem: despite rising investment in AI, companies are failing to capture up to 40% of the productivity gains the technology could deliver, because of shortfalls in talent investment. Her explanation is that many training programs focus on technical proficiency with AI tools and leave aside the critical skills — judgment, evaluation and oversight — that actually determine whether an AI strategy ends up paying off.

On what AI still cannot do, the article argues that the technology excels at repetitive tasks and at surfacing data, but that it falls to people to define the goal, set the purpose and provide the context that gives that information meaning. It is people who decide which objectives are worth pursuing, who spot when the context changes and who remain accountable for the results. Without that human direction, the author maintains, data and insights may not be applied effectively or generate their full potential value. From this perspective, an AI-orchestrated workplace does not eliminate the person, but rather the parts of the job that never required human judgment in the first place; what remains is precisely the part only a human can do.

The article argues that "AI fluency" should be understood as a leadership skill and not as an isolated technical specialty: it combines knowing how to operate the tools with assessing how relevant their outputs are and knowing how and when to step in if something goes wrong. To illustrate the scale of the challenge, the author cites a striking figure: within the average organization, machine identities outnumber human employees by 82 to 1. Far from diluting human responsibility, she argues, that scale concentrates it: the more autonomous AI systems become, the more important human judgment becomes. On this point, she mentions a World Economic Forum survey according to which analytical thinking is the skill employers most look for when hiring, with roughly seven in ten companies rating it as essential. The conclusion the author draws is clear: use AI to make work more efficient, not to make the judgment calls that remain the most important part of the job.

Another of the article's central themes is that no organization solves the integration of AI into its workforce on its own; going it alone consumes time and resources because companies end up rebuilding processes that others have already solved. As an example of collaboration with ecosystem partners, the author cites EY's work with beverage manufacturer Lion, which achieved a 75% faster customer response and an improvement of roughly 30% in the operating costs of its people function by leaning on partners that had already solved parts of that problem in other contexts, instead of building every layer of an AI-enabled people function from scratch.

The piece also makes the case for redesigning workflows so that people remain at the center of decision-making. It criticizes models based on rigid handoffs — in which AI completes a task before passing the result to a person — because they assume tasks can be split cleanly, something that in practice reduces the opportunities for meaningful judgment and can create more iterative work. Instead, it proposes a more durable model organized around shared tasks, with a genuinely integrated process in which humans and AI contribute at every stage rather than sequentially. As an illustration, the article mentions the case of Daikin, the multinational specializing in climate control, which combined human and AI teams working in tandem to roll out a new ERP system: AI assisted with code generation and automated testing, while employees retained oversight of exceptions and high-risk scenarios. This approach, it is stated, accelerated delivery by roughly 30%, and the pilot project produced a 10% improvement in counter efficiency and a 20% faster financial close.

The article closes with the idea that, given the constant evolution of AI models, technical skills tied to a specific version of a model have a short shelf life, whereas the judgment to know when and how to intervene does not expire. That judgment, the author stresses, is essential to building safe and trustworthy AI, and it does not happen by accident: it requires fluency at every level of the organization, partners who share the burden of building those capabilities, and workflows designed to keep people involved at every stage with the right oversight. Her conclusion is that the organizations getting AI at scale right are those making deliberate decisions about which parts of the work should remain human, and that AI adoption done well clarifies that human role rather than shrinking it.