Enterprise AI buying is beginning to diverge from the rest of the technology stack. Companies are concentrating deployments around fewer use cases, but unlike cloud, security, analytics and other IT categories, they are still planning to bring more AI suppliers into the mix.
Boston Consulting Group (BCG), a management consulting firm, reported the pattern in its latest “IT Spending Pulse: AI Takes Priority as Confidence Returns,” published Sept. 17. The survey covered 423 technology decision-makers at director level or above at midsize and large organizations. Of every category BCG tracks, AI and machine learning (ML) was the only one in which more buyers planned to expand their supplier roster than consolidate it over the next year.
The difference was substantial. BCG found 55% of buyers planned to add AI/ML suppliers, compared with 21% that planned to consolidate — a difference of 34 percentage points. In cloud, the two were evenly balanced, while every other category leaned toward consolidation.
Yet respondents at every level of AI maturity were pursuing fewer generative AI (GenAI) and AI agent use cases on average than in BCG’s mid-2025 survey.
Adoption increased in analytics, customer service, security and risk prevention, and internal communications, even as the average number of use cases fell overall. BCG describes the shift as a move from broad experimentation to more disciplined execution, but it does not explain why buyers also plan to add more AI suppliers.
Supplier choice remains open
Separate 2026 research suggests new vendors remain firmly in the consideration set. Info-Tech Research Group, an IT research and advisory firm, found in its “AI Adoption and Impact Study: AI in the Enterprise” that 80% of 522 respondents to its sourcing question preferred buying AI capabilities rather than building them in-house. But they diverge sharply on who they want to buy from: 42% of all respondents favored AI from existing vendors, while 38% preferred new, best-of-breed suppliers.
Info-Tech’s survey does not explain BCG’s finding, but it does validate that incumbents don’t have a lock on the next AI purchase.
Returns rise with maturity
BCG classed companies with AI embedded in enterprise strategy and scaled across multiple functions as high-maturity adopters. Those firms reported roughly 19% weighted-average return on investment (ROI), compared with 8% to 9% among less-mature adopters.
Average ROI on GenAI and AI agent deployments sat at 13.8%, up from 11.2% in BCG’s mid-2025 survey. BCG said much of the overall increase reflected a larger share of respondents in higher-maturity groups.
Info-Tech principal research director Brian Jackson framed the same shift in a release accompanying that company’s study: “Enterprise AI is moving past the question of whether organizations should experiment and into the question of how they prove value.”
Mature buyers loosen the reins
BCG also found higher-maturity firms had better visibility into AI and token costs, often down to the team or user level. Those firms tended to use more permissive token policies, while less-mature respondents leaned toward harder limits and more cautious controls.
BCG said greater visibility allowed mature firms to give heavy users more freedom without a proportional increase in spending. Power users, defined as the top 10% of AI adopters, generated outsized productivity without consuming a disproportionate share of AI spending, according to the report.
Neither survey establishes whether buyers’ plans to add AI suppliers will persist as the technology matures. For now, the discipline BCG describes is showing up in what companies deploy, not in whom they buy from — which leaves the next AI purchase open to whomever makes the best case, incumbent or not.