The bill is coming
Skill atrophy and AI subsidies are making organisations dependent on OpenAI, Anthropic and few others. The bill is coming.
Both companies are running at a loss. Investor money covers the gap between what customers pay and what these services actually cost to run. This is deliberate — buy market share first, sort out the economics later.
The problem is that later eventually arrives. If they fail to reach profitability, prices go up or the service disappears. Organisations that built their operations around current pricing are making a bet they probably never consciously decided to make.
The deeper issue is what happens to people in the meantime.
When organisations adopt AI to replace rather than augment, institutional knowledge leaves with the people. This is not a hypothetical. It follows directly from how most cost reduction programmes work — you cut headcount, the remaining people use the tools, and nobody invests in building new skills because the tools make that seem unnecessary.
Then pricing changes. Or the service degrades. Or integration breaks. And you need capabilities that quietly disappeared two years ago.
Responsible AI adoption means two things. Know which capabilities cannot be allowed to atrophy, and protect them deliberately. Build the ability to switch providers before you need to, not after.
The organisations that treat AI as leverage will have options. The ones that treat it as a replacement will find out what dependency actually costs.