A company can buy AI tools from different vendors and still depend on the same underlying models, APIs or computing providers. Swati Deepak Kumar, a Citi senior vice-president of IT, argues that this can leave important business decisions exposed to the same blind spot or behaviour change, even while every system appears to be running normally.
Watch Desk analysis
What happened
In a Forbes Technology Council essay, Kumar calls this risk “cognitive concentration”. She presents it as a management lens, not an established regulatory term: the concern is that a change or flaw in a shared AI dependency could affect several workflows at once.
Kumar cites a 2024 Bank of England and FCA survey in which the three largest third-party providers accounted for 44% of reported model providers at financial firms. She also cites a 2026 IBM Institute for Business Value survey: 91% of surveyed executives said they did not fully understand their AI dependencies, while 71% said switching their primary AI vendor or model would be difficult.
Why it matters
A shared dependency may not cause an obvious outage. Instead, separate systems could continue operating while their judgements shift in related ways. Kumar says the risk deserves most attention where workflows are consequential and hard to replace, such as lending, cybersecurity or production changes.
Her practical advice is to map model families, API and computing providers, rank workflows by consequence, and test how long switching actually takes. A backup contract is not much of a fallback if nobody has tried using it.
Our read
This is a useful way to look beneath a tidy vendor list. The point is not to collect models like spare umbrellas, but to know where dependencies overlap and whether the business can cope if one becomes unreliable. Kumar’s essay makes an argument for resilience, not proof that shared models have already caused widespread correlated failures.
What to watch
- Whether companies start mapping model and infrastructure dependencies across critical applications.
- Whether fallback plans are tested, rather than merely written into contracts.
- Whether major model changes are staged and checked before reaching multiple workflows.
Discussion spark: Should businesses prioritise using independent AI models in their most consequential workflows, or is a tested fallback enough?
Sources and evidence
- Cognitive Concentration: The Hidden Risk Beneath Enterprise AI – Forbes (6 October 2026, 14:30 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.