The week in brief — AI's real barrier is never the technology, it's trust and what you do with it

Four cards this week, one thread: what blocks or derails AI is almost never the technology. Handing a decision to a third party's AI doesn't hand off responsibility (an HR screening tool accused of filtering out over-40 applicants, Mobley v. Workday, 2025); a demo proves nothing until the AI is measured on your own cases — the “evals” that would grow from 18% of engineering teams in 2025 to 60% by 2028 (Gartner). On the agent side, nearly two-thirds of organizations name security and risk, not technical limits, as the top barrier to scaling (McKinsey, 2026), while 43% of data breaches already involve unvetted “shadow” AI that two-thirds of companies fail to govern (IBM, 2026). The takeaway fits in a sentence: any SME or mid-sized firm is better off investing less in the tool and more in what surrounds it — proof on real cases, owned responsibility, and written verification and usage rules before the first deployment.

Source: YPrtnrs cards of the week (Mobley v. Workday, 2025; Gartner, 2025; McKinsey, “State of AI trust in 2026”; IBM, “2026 Cost of a Data Breach Report”).