87% of large companies say nothing about what AI does to their workforce
The silence in AI disclosures is a reporting gap that boards can close before regulators and investors close it for them.
Sign up for my Substack daily AI newsletter https://johanosteyn.substack.com
Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/
Most AI business cases I have seen have a column for the people they will no longer need. Very few have a column for the people they will help, the roles they will change or the work they will create. At a recent public lecture at Nelson Mandela University, Nkemdilim Uwaje Begho, chief executive of Futuresoft, put the problem in one line: "We must have metrics that are people first and not automation first". Boards should take that as a practical instruction, not a slogan.
CONTEXT AND BACKGROUND
The silence is measurable. The World Benchmarking Alliance examined the disclosures of 200 large companies in July. While 39% now publish AI principles, only 13% mention the workforce in an AI context, 6.5% link AI to reskilling, and a single company, Deutsche Telekom, states plainly that AI may displace jobs. Some 87% make no reference to workforce impact at all.
Meanwhile, the cuts are being announced. In the United States, employers named AI as the leading reason for job cuts for three months running, and it accounted for 40% of announced cuts in May. African technology companies cut a record 2,574 jobs in the first half of 2026, with AI among the drivers.
In South Africa, the stakes are higher than anywhere those figures come from. The official unemployment rate rose to 33.6% in the second quarter, and 47.4% among people aged 15 to 34.
INSIGHT AND ANALYSIS
A business case that counts only heads removed is not neutral. It makes savings visible and everything else invisible, so the board approves what it can see. Saker Nusseibeh of Federated Hermes made the investor version of this point at the IFRS Foundation conference in June, noting that "the trained people" behind a company's productivity are largely invisible in its accounts, and welcoming the ISSB's research on human capital disclosure.
I have previously written about this when the employment data began contradicting itself, arguing that AI expands well-resourced firms and substitutes for workers in others. A second column is how a board finds out which kind of firm it is running.
The pressure will not stay voluntary. Deputy President Mashatile has asked labour federations to draft a Digital Workers' Charter including "binding transition plans with impact assessments" and consultation with workers on automation decisions. Companies that already measure the effect of AI on their people will negotiate from evidence. The rest will negotiate from silence.
The fair objections deserve answers. Companies exist to make returns, and job creation is primarily the state's task. Attribution is hard, since announced reasons for layoffs are what employers say rather than independent measurement. A "jobs created" figure can become a vanity metric. Planned job changes can be price-sensitive for listed companies, and some firms already report reskilling spend elsewhere in their sustainability reporting. Each has some truth, and none argues for measuring nothing. The point is not to promise employment or announce restructuring early, but for the board itself to see the human consequences of a decision before it is taken, and to report actual outcomes afterwards.
IMPLICATIONS
Boards can act at the next AI approval. Ask for a second column with three measures: roles changed and how, people redeployed or reskilled with funding attached, and new roles created inside the company or among suppliers. Set targets only where they are credible, and report actuals in the integrated report beside the savings.
Audit and remuneration committees should treat these numbers with the same scepticism as financial ones. If executives are rewarded for cost reduction alone, the second column will stay empty.
There is a practical reason as well as a moral one. Retrenchment under section 189 of the Labour Relations Act already requires consultation on ways to avoid dismissals, so the questions will be asked in any case. A company that has tracked redeployment and reskilling can answer them with records rather than intentions. Smaller firms need not build a reporting system; a single page in the business case, reviewed a year later, would already put a company ahead of the 87% of large companies that disclose nothing.
CLOSING TAKEAWAY
AI will remove some work, and pretending otherwise helps nobody. The question for a South African board is whether it can show, with numbers, what its AI built as well as what it cut. In an economy where a third of the labour force is officially unemployed, that is the evidence on which the social licence to automate will be granted or withdrawn.
Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence's impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan's work at https://www.aiforbusiness.net
Comments