A new paper from Boston Consulting Group argues that agentic AI is about to double the value delivered by AI across the economy. According to the study, a small cadre of 5% of businesses have achieved ‘future built’ status to enjoy 6% revenue growth on the back of agentic AI – showing the way in which others could finally see returns on investment into the technology.
Sentiment around artificial intelligence has shifted noticeably in 2025 – with businesses and investors increasingly determined to see a return on the substantial amounts of capital they have invested so far. The majority of AI reportage might still focus on breathless estimations of trillion-dollar productivity boosts, or a mass replacement of human talent with digital tools, hushed whispers around whether the promise of the technology will become reality before the money runs out have steadily gone mainstream.
Amid this, the consulting sector has come under notable pressure – having sold many clients on the apparently endless opportunity presented by AI. Following other digital hypes, including blockchain, the metaverse and NFTs, however, their assertion that eventually large language models will come good have not entirely been taken at face value. So, at a time when consultants are pushing for companies to plumb AI into their organisations, the flat growth of the consulting sector suggests they are either being ignored, or less trusted on the lucrative implementation that would come with such a change.
Boston Consulting Group has sought to turn the tide back on this basis, with the release of a new study claiming that if ‘laggards’ who have been slow or ignored AI would adopt the practices of a small number of early movers on ‘agentic AI’, they could enjoy meteoric growth before the end of the decade. And amid a sustained global economic slump, those promises of lines pointing dramatically upward in relation to a fresher, less familiar form of AI might serve as something of a siren’s call to clients.
Unlike traditional models, the proponents of agentic AI claim the systems can ‘learn’, ‘reason’, and ‘act autonomously’ to ‘solve’ complex, multistep problems. In theory, this enables a model to refine its decision-making over time, leading to improved outcomes.
Polling 1,250 global business figures, BCG determined a select group of 5% of firms represented in the pool were ‘future-built’, due in large part to their practices around agentic AI. Compared to ‘laggards’, who saw a revenue increase of 1.2% following AI implementations, these leaders enjoyed a 6.2% increase, BCG found – while they also saw 4% higher levels of cost reduction (relating directly to gains where AI was applied, rather than across the whole organisation). By 2028, these future-built firms could even enjoy revenue spikes of 14.2%, and cost reductions of 13.6%, BCG argued.
“The technology is advancing weekly, and leading companies are accelerating,” said Michael Grebe , a managing director and senior partner at BCG and a co-author of the report. “For the majority of firms, catching up will require more than investment: it will take reinvention. The good news is that the playbook followed by future-built companies is clearly delineated and available to all.”
The findings could lead to a growing gap between the two camps, according to the researchers – unless the other 95% of companies follow the road-map established by future-built firms. BCG noted the leading cohort of companies allocated 15% of their AI budgets to agents. A third of these companies used agents, compared with 12% of companies that are only scaling AI, and almost none of the other 60%. To move up the scale, the strategy consultancy suggested firms should look to focus on achieving a “concreate path to value”, and once they have generated these complex plans, they should look to scale their “AI ecosystem with long-term partnerships” – which might or might not include sourcing external expertise and advice.
It might not fill companies with confidence that the 5% figure mirrors the percentage of firms MIT’s infamous paper recently suggested had seen any return at all from AI investments. The research found 5% of organisations which had successfully scaled AI tools into production were said to be “extracting millions in value”. In particular, Aditya Challapally, the MIT researcher who led the study, told Fortune that some large companies and younger start-ups were “excelling” with AI because “they pick one pain point, execute well, and partner smartly with companies who use their tools.” That reportedly led some start-ups led by young founders to see revenue “jump from zero to $20 million in a year”. That might not translate elsewhere, though.
Outside of a carefully controlled, highly curated environment, the jury is still well and truly out for agentic AI. While complex WebArena web-task benchmarks have estimated accuracy rates of agentic AI between 40% and 60% at different points in the year as a sign of progress, deploying a technology still liable to be wrong about things almost half the time would be catastrophic for most large companies. This could mean that while small challengers might make fast money from AI, the larger they grow, the less effective their tools may become – while firms which are already large, complex organisations will face huge hurdles in privacy, security, regulatory compliance and accountability, if and when they deploy agentic AI in the near-future.
