Boston Consulting Group

Jens Kengelbach's team counted 1,355 AI-related acquisitions in 2025, 53% of them by non-tech buyers, plus 1,340 AI alliances and joint ventures, more than double the 2024 figure. Boston Consulting Group reads that market through a single grid: differentiation against availability on the market, which sorts a capability into adopt, partner, build, or buy. The guiding principle is the lightest structure that gives the needed access or control while keeping the freedom to switch. Its own measurement complicates that: about nine industry-tech partnerships in ten miss their objectives, and a light structure is not simpler for it.

Investors pay a premium for the trade: announcement returns above 0.8% when a non-tech company buys AI software, roughly three times the average. Boston Consulting Group's position is that for a non-AI company, the advantage lies in the data, workflows, IP, and expertise around the model rather than in the model itself, and that as models and tools commoditize, that is where it will stay.

A second line of work turns inward. Six of its researchers, including Gabriella Rosen Kellerman, surveyed 1,488 full-time US employees in January 2026 and reported in Harvard Business Review that AI can cut burnout by 15% when it absorbs repetitive toil while raising cognitive fatigue where oversight is intensive: +14% mental effort, +19% information overload. Earlier, Ethan Mollick's research with the firm, 758 consultants, produced the jagged-frontier result: +40% quality inside the frontier, 60-70% accuracy outside it against 84% unassisted.