Why 42% of companies abandoned their AI initiatives (and the hire that predicts survival)

Companies are not failing to adopt AI. They are failing to keep it. S&P Global's 2025 enterprise survey found 42% of companies had abandoned the majority of their AI initiatives before they reached production, up from 17% a year earlier. The same survey put the average attrition at 46% of projects scrapped between proof of concept and broad adoption.

The direction was predicted. Gartner said in July 2024 that at least 30% of generative AI projects would be abandoned after proof of concept by the end of 2025, naming poor data quality, inadequate risk controls, escalating costs and unclear business value. MIT's NANDA project (The GenAI Divide, 2025) reported 95% of organisations investing in generative AI seeing zero return; the study is preliminary and its sample small, but it points the same way as the measured surveys.

Picture the program review where the abandonment decision gets made. The pilots worked. The vendor demos were real. What is missing from the room is a single person whose number moves if the pilots never reach the P&L, and programs without that person become the statistics above.

BCG's 10-20-70 principle describes where successful AI leaders actually put resources: 10% into algorithms, 20% into technology and data, 70% into people and processes. In other words, for every dollar spent on models, expect seven on the humans and workflows around them. Abandoned programs almost always inverted that ratio.

For every dollar spent on models, expect seven on the humans and workflows around them.

The clearest measured difference between programs that survive and programs that join the abandonment numbers is an accountable AI leader. IBM's study of more than 600 chief AI officers found organisations with the seat see 10% higher return on AI spend, rising to 36% where the leader runs a centralised or hub-and-spoke operating model, and 61% of those leaders control their organisation's AI budget.

Boards have noticed. IBM's 2026 CEO study reports 76% of organisations now have a chief AI officer, up from 26% a year earlier. Accountability cuts both ways, though: Gartner predicts that by 2027, 75% of chief data and analytics officers not seen as essential to their organisation's AI success will lose the C-level position. Boards are not buying more pilots. They are buying someone to own the number.

Boards are not buying more pilots. They are buying someone to own the number.

Australia is hiring into the same correction. Galileo's July 2026 tracking counted 53 live AI leadership ads across the country carrying 38 distinct titles, and the largest cluster, 30 of 53, were adoption mandates: leaders hired to turn existing pilots into used systems rather than to build new ones. The demand side has already diagnosed the problem in the abandonment data.

The supply side is thin and passive. Leaders at this level rarely apply to ads, and roughly nine in ten live Australian AI postings hide pay (Galileo tracking, 496 postings, July 2026), which suppresses the response from the candidates who do look. LinkedIn's Jobs on the Rise 2026, reported by ACS Information Age, ranks AI engineering Australia's fastest-growing job, so the team behind the leader is getting more contested each quarter as well.

Before restarting a stalled program, settle three things. Name the accountable owner and decide the mandate type first: a builder, an adoption leader or an enterprise owner, because the three profiles fail in each other's seats. Fund the team with the leader, since the first placement is typically the first of about five seats over 12 months and an unfunded leader spends year one negotiating instead of delivering.

Then set the number: the one figure the board sees each quarter, agreed before the leader signs. Programs with an owner, a funded team and a number are the ones on the right side of the 42%.

Galileo Search places the accountable AI leaders Australian boards are hiring, Head of AI, Chief AI Officer and AI Transformation Lead, plus the teams they build, grounded in live demand tracking and more than 22,000 salary data points. Restarting a stalled program? Start with the leader brief and we will pressure-test it against the market.