AI is becoming an expectation at work. But what does it mean in practice?

A public institution rather than a company, the European Commission contributes evidence of deployed training segmented by audience. Its AI@EC learning packages are oriented to generalists, managers and developers, supported by an internal AI web portal, with trainings categorised as essential, highly recommended or recommended. The managers package covers governance topics such as AI rules of engagement and AI Act impact.

No company in this evidence base pairs executive intent with role-level detail as closely as DBS. Its leadership signals treat AI as a workforce matter: 13,000 staff earmarked for upskilling, a dedicated academy for its 5,000-strong technology team, and a stated view of how AI changes contract roles.

Unusually for this evidence base, EY contributes a publicly verifiable credential standard rather than a training programme. Its AI badges set required tiers for consulting and advisory staff: Bronze demands 15 hours of training, Silver hands-on practice in one or two AI capabilities, and Platinum knowledge and oversight in a large-scale AI programme, all verifiable on Credly.

Few organisations here state their targets as plainly as Ingka Group. A 30-minute Say Hej to AI e-learning covers all co-workers, with a stated aim of 100% basic literacy across the workforce by 2027; 649 senior leaders have engaged in ethical AI sessions, and roles more likely to involve high-risk AI use receive specialised training.

Building, not merely using, is the yardstick in Ramp's evidence: a four-level AI-native ladder signalled for all employees. Competent users tweak GPTs and internal tools, non-technical builders ship apps that automate part of their job and can commit code with AI, and systems builders create AI infrastructure that accelerates the whole team, while the disengaged are bluntly told they are unlikely to last.

Breadth and role specificity combine in JPMorgan's leadership signals. LLM Suite, its proprietary internal assistant, reached 200,000 users within eight months of its early 2024 launch, AI Made Easy has involved tens of thousands of staff, and the stated objective is to train every employee on AI for their specific role, with software engineers upskilled to build agents.

Efficiency numbers anchor Amazon's leadership signals. Andy Jassy's June 2025 memo told employees the corporate workforce is expected to shrink as agents and generative AI take on more work, while Amazon Q Developer is credited with saving an estimated 4,500 developer-years and $260m in annualised gains from Java-modernisation work alone.

Mandatory requirements and live deployment intersect in Citigroup's evidence. AI literacy and prompt-engineering training is compulsory for all staff, around 180,000 employees in 83 countries had AI-tool access by Q3 2025, and more than 2,000 AI Champions support adoption, while agentic AI has been deployed to 5,000 employees simultaneously across four countries for real work rather than a pilot.

Consequences distinguish Accenture's leadership signals from most training claims. More than 550,000 of its roughly 780,000 people have been trained in the fundamentals of generative AI, promotions are tied to demonstrated AI proficiency from FY26, and Julie Sweet has said the firm will exit staff who cannot be reskilled on AI.

Reflexive use is the standard Shopify's evidence sets. Tobi Lutke's memo of 7 April 2025 made reflexive AI usage a baseline expectation for every employee, prototyping now happens with AI in the GSD phase, and AI questions have been added to performance and peer reviews, so falling short of the baseline is visible in formal evaluation.

Thirteen role families, each with four defined fluency levels, make Zapier's published hiring rubric the most granular standard here. From support and legal to engineering and finance, the required bar rises from using AI to operate at a meaningfully higher level, through orchestrating systems, to re-engineering how work happens, with level-zero behaviours spelt out just as concretely.