Why corporate Australia is grading employees on a skill it can’t evaluate

Australian enterprises are eagerly grading and promoting employees based on their technical capabilities, yet they are leaving middle managers completely unequipped to evaluate what good or bad AI usage actually looks like. This contradiction exposes a flaw in how organisations are making talent decisions, and it has direct consequences, like inconsistent promotions and performance ratings that don’t reflect actual capability. The result is AI performance theatre, where the appearance of AI tool adoption is rewarded over productivity gains, while the people doing the harder work of redesigning how the businesses operates go unrecognised. 

It’s no surprise how we got here with the AI-productivity race hot on our heels, so let’s unpack the hazards of moving at speed and how we can equip middle managers with the skills to end up on the winning team.

The scramble to metricise technical behaviours

The aggressive shift toward institutionalising AI proficiency is already well underway. Our latest study into AI & workforce management reveals that 74% of Australian organisations have explicitly linked AI skills to promotion criteria, while over half are directly tying these technical capabilities to performance ratings.

We are hastily updating traditional HR rubrics with broad, ambiguous mandates, like expecting staff across non-technical functions to demonstrate efficient use of GenAI tools, without mapping what that looks like in practice. For instance, what constitutes strong AI skills for a financial auditor looks different from those of a marketing coordinator. Recent Australian research into AI integration highlights this as the "automation-augmentation paradox", warning that organisations must carefully balance efficiency gains of AI with maintaining human judgement and decision-making. When we fail to treat AI literacy as a clearly defined organisational competency, we lose this critical balance.

Therefore, the most critical competencies will revolve around nuanced behaviours like workflow evaluation, proactive output quality verification, and handling sensitive data appropriately. Yet, we are holding employees accountable for something that the business itself hasn't clearly defined.

Blind optimism and the evaluation illusion

This exposes a dangerous contradiction between executive confidence and the total lack of middle management enablement. While 81% of leaders confidently agree that their managers can reliably judge an employee’s AI proficiency in performance reviews, the reality on the ground tells a very different story. Just 24% of organisations actually provide manager enablement training on assessing and coaching employees on effective AI use.

Middle manager capability is the single biggest lever determining which side of that gap an organisation lands on, yet it’s the one most businesses haven’t invested in.

This dynamic misaligns the internal talent pipeline, penalising the diligent and rewarding the performative. While AI-enabled operations can unlock 2.4 times greater productivity, only 15% of organisations are actually redesigning how work is performed to capture this value. Realising the full potential of AI requires rethinking operating models, designing scalable talent systems, and embedding human accountability. In fact, if half of all beginner-level workers improved to intermediate levels, the total dividend would be $18.9 billion for the greater economy, according to RMIT Online.

If organisations fail to equip their managers to assess the competencies, every promotion, pay decision, and performance rating built on AI proficiency is effectively a guess, and that undermines the credibility of every talent decision that follows. 

Moving beyond the theatre

If we want to build a workforce that can actually use AI well, we have to stop blindly trusting the process and start actively equipping the managers we rely on to evaluate it. To move beyond performance theatre, HR leaders must implement three critical steps:

1. Standardise plain-language baselines: Establish clear, practical standards for what strong AI skills look like across different business units. The focus must shift away from simply “using” to high-value skills like auditing output quality, designing efficient human-AI handoffs, and protecting sensitive data.

2. Empower the evaluators first: Pause or recalibrate AI performance mandates until line managers receive comprehensive training. If an organisation expects a manager to gatekeep promotions based on AI proficiency, managers must first be taught how to audit workflows, spot bias, and provide constructive technical coaching.

3. Reward value, not volume: We must evaluate the actual business outcomes an employee delivers. Stop rewarding the sheer velocity of AI-generated deliverables. Instead, credit employees for the resources they recoup, the time they save, the risks they mitigate, and the processes they genuinely improve.

AI transformation cannot be achieved by dictating KPIs from the top and hoping for the best in the middle. If performance evaluation becomes a guessing game, workplace equity is the first casualty. True corporate transformation requires genuine capability. It is time for businesses to train the judges, not just test the players.

Anna Volkova is the head of people and culture at HiBob APJ