The Concentration Trap

AI is not a rising tide lifting all firms; three-quarters of the gains accrue to the disciplined fifth that redesigned how they work.

Sign up for my Substack daily AI newsletter here.

See my AI Training course portfolio for corporate Business Leaders here.

Article link: https://open.substack.com/pub/johanosteyn/p/the-concentration-trap?r=73gqa&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/

There is a comfortable assumption in many boardrooms that buying AI is the same as modernising with it. Write the cheques for the enterprise licences, the thinking goes, and the company joins the future. The evidence points somewhere more uncomfortable. AI is not spreading its rewards evenly across the market. It is widening the distance between a disciplined minority and everyone else.

CONTEXT AND BACKGROUND

The clearest sign is in where the value lands. PwC’s 2026 analysis found that roughly three-quarters of AI’s economic gains are accruing to about a fifth of companies, the firms treating AI as a lever for growth rather than a way to shave costs. This is not a rising tide. It is a concentration of advantage in the hands of organisations that have done the harder work of rebuilding how they operate, while the rest buy the same tools and watch their margins stay flat. The technology that looks broadly available is quietly sorting firms by capability.

INSIGHT AND ANALYSIS

What separates the winning fifth is not budget but architecture. They do not treat AI as an application on an employee’s desktop; they treat it as an infrastructure layer wired into core operations, with integrated data pipelines and governed model choice rather than tools bolted onto processes that never changed. The laggards adopt; the leaders redesign. That distinction compounds, because redesigned workflows generate value that funds further redesign, so the gap widens with each cycle rather than closing. A firm that merely buys licences is paying a dollar-denominated cost while capturing little of the structural efficiency that is meant to offset it, and the leaders are using those same tools to lower their cost-to-serve and undercut it.

For South Africa, the trap is especially live, because the local market sits disproportionately on the wrong side of the divide. PwC’s Africa CEO survey found that businesses remain stuck in experimental phases, struggling to move pilots into enterprise-wide deployment, with only 41 percent of CEOs holding a clear AI roadmap and just 8 percent willing to make large investments, even as 81 percent express optimism well above the global average. Optimism is not a strategy. A promising pilot that cannot scale across the business because of fragmented data and risk-averse investment is just an expensive proof of concept, and PwC’s own warning is blunt: those investing modestly today risk falling behind competitors scaling rapidly.

I have previously written about this concentration at the level of the wider economy, in a piece on the rise of the digital oligarchy, where a narrow elite controls the platforms and infrastructure that everyone else rents, leaving countries that consume rather than create these technologies with little say. The same dynamic now plays out between firms in a single market, and the firms on the wrong side of it are funding their own relative decline.

IMPLICATIONS

This makes AI a competitiveness question, not a procurement one, and that belongs to the board. In a slower-growth economy, a company that fails to convert AI spend into operating advantage does not simply miss an upside; it loses ground to better-run rivals that can underprice it, out-serve it and attract its talent. The honest diagnostic for a board is uncomfortable: are our AI projects isolated experiments or an enterprise capability anchored in data that can actually scale, and can we point to a single workflow we rebuilt rather than merely a set of seats we bought. A firm that cannot answer is most likely in the majority that pays and waits, not the minority that compounds.

CLOSING TAKEAWAY

The reassuring story is that AI democratises productivity. The evidence says it rewards the discipline a firm already has, which means it can deepen the gap between leaders and laggards rather than close it. The way out of the trap is not more licences but the unglamorous work the winning fifth have done: redesigning workflows, governing how data and models are used, measuring value at the level of the process, and training managers rather than merely handing tools to users. For a South African market already scaling slowly, the cost of staying in the experimental phase is no longer standing still. It is falling steadily further behind.

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