CIOs Should Make AI Learning Time a Fixed Part of Working Hours

AI literacy is entering performance reviews and job interviews. Yet only 24 percent of surveyed employees have working-time rules that include learning, although Article 4 of the AI Act requires deployers to support its development. Relegating AI experiments to employees’ own time shifts a management responsibility onto staff.

RelatedGoogle Pays Publishers for AI Answers

Huntley’s Argument and the Objection

On 9 October, developer Geoffrey Huntley published “replaced”. He argues that AI use is no longer optional in software careers. He advises hiring managers to schedule development time and assess employees’ AI literacy. He distinguishes producers, who automate workflows with AI, from consumers, who simply use the tools. He leaves the assessment methodology unspecified.

Software engineer Addy Osmani responded to Huntley’s post with an objection that shifts the debate toward organizational responsibility. A team allowed to experiment with AI only after hours receives an implicit message that its use does not count. This raises the question of when employees should develop the required skills.

What distinguishes producers from consumers? Developer Geoffrey Huntley proposes this distinction to assess employees’ AI literacy. Producers automate workflows with AI; consumers simply use the tools. No assessment methodology exists, so companies must define their own criteria.

Willingness to Learn Meets Working Hours

Among employees surveyed for Bitkom Akademie’s 2026 continuing education study, 73 percent use AI at work, up from 64 percent in 2025. 79 percent want to improve their AI literacy. Yet only 24 percent have working-time rules that include learning. Only 21 percent also report that their manager actively raises professional development.

A contradiction emerges at management level. To address changing skill requirements, 77 percent of surveyed executives primarily favor training existing employees over automation and recruitment. 45 percent acknowledge that investment favors technology over training. 51 percent have embedded professional development in their strategy.

The gap is therefore organizational. Leaders favor training but invest more in technology and rarely schedule learning time. Companies also cite time constraints in a representative Bitkom survey conducted in March among 604 companies with at least 20 employees. 41 percent use AI, up from 17 percent the previous year. 66 percent cite insufficient time as a barrier to digitalization.

What Article 4 Requires

Article 4 of the AI Act has applied since 2 February 2025. It was revised by Omnibus Regulation (EU) 2026/1744, effective since 27 July 2026. It requires providers and deployers to take measures supporting their staff’s AI literacy development. Deployers are companies using AI systems. They are explicitly exempt from guaranteeing any individual’s specific level of competence.

According to the Federal Network Agency (Bundesnetzagentur), this obligation covers deployers of all sizes, including those using general-purpose chatbots. Insufficient AI literacy can constitute a breach of the duty of care, particularly when harm occurs.

Including Learning Time in Capacity Planning

The first lever for CIOs is a fixed time allocation for experimentation. This belongs in capacity planning: allocating available team hours to operations, projects and further development. Huntley’s advice to schedule development time points in the same direction. Following Osmani’s logic, this allocation sends the opposite signal: AI use counts. It also constitutes a concrete measure of the kind Article 4 requires from deployers.

Rethinking Assessment and Hiring

The second lever concerns performance reviews. Distinguishing producers from consumers offers an assessment standard beyond mere use. Without a methodology, IT leaders must define their own criteria, such as which automated workflows qualify as documentary proof. This standard remains fair only if schedules allow time to progress from consumer to producer.

The third lever is the interview guide. Huntley identifies opposition to AI, covert AI use during interviews and a lack of firsthand experience as warning signs. His baseline is being able to explain at a whiteboard how a coding agent works with a context window and tokenization. A coding agent is an AI tool that independently writes and modifies code. The context window is the amount of text a language model processes at once; tokenization splits text into small computational units. He considers personal projects a strong indicator, and verifiable organizational changes for managers.

A purely AI-first label that makes AI affinity the primary hiring criterion has a downside. In the thread on Huntley’s post, the account @bertlef_ warns that hiring only AI-oriented people could harm a team because nobody would solve problems without tools.

Michael Taiwo commented on the provocative recommendation to quit if an employer bans AI. AI literacy matters, he argued, but quitting is costly advice for someone paying rent for three people.

Frequently Asked Questions

More from the MBF Media Network

Header image source: AI-generated (October 2026)

Translated from the German original using artificial intelligence. The German version is authoritative.