AI adoption scales rapidly but readiness lags, says Capgemini Research Institute

 Enterprise adoption of generative AI (Gen AI) has grown at breakneck speed, with usage expanding fivefold in just two years, according to the Capgemini Research Institute’s latest report, Harnessing the value of AI: Unlocking scalable advantage. The study finds that nearly six in ten organisations expect AI to function as an active team member or supervisor for other AI systems within the next year, compared to 44% today.

Despite this surge, most firms admit they are unprepared for such close human-AI collaboration. Two-thirds of organisations say they will need to restructure teams to support new modes of interaction between employees and AI systems.

From experimentation to scale

The report highlights that 30% of enterprises are now fully or partially scaling Gen AI, up from just 6% in 2023. Today, 93% are either piloting or enabling Gen AI across business functions, with telecom, consumer products, and aerospace among the frontrunners. Customer operations, marketing, risk management, and IT remain the most common areas of implementation.

“Enterprise adoption of AI is scaling faster than almost any technology we’ve seen before, but rapid adoption doesn’t always mean large-scale deployment with measurable returns,” said Franck Greverie, Chief Technology & Portfolio Officer at Capgemini. He emphasised the need for trusted, secure data foundations and balanced human-AI operating models.

Rising investments and cloud costs

Organisations are backing their ambitions with capital. Nearly nine in ten firms increased Gen AI spending in the past year, dedicating on average 12% of their IT budgets to the technology. More than 60% expect further increases next year.

But the rush has also triggered unexpected costs. Over half of enterprises reported “bill shocks” from cloud consumption as AI pilots scaled faster than projected. To contain costs, many are exploring small language models (SLMs) as alternatives to large AI systems.

The rise of AI agents and multi-agent systems

AI agents, software systems that handle business tasks independently, are gaining traction. Most executives in functions like R&D, sales, and marketing believe agents will manage at least one process in the next three to five years.

Of those scaling AI agents, 45% are already piloting multi-agent systems, where interconnected agents coordinate tasks. Nearly four in ten leaders expect these systems to evolve into self-learning agents needing minimal human supervision.

Yet trust remains a hurdle: 71% of organisations said they cannot fully trust autonomous AI agents for enterprise use. Governance frameworks are still catching up, with less than half of organisations having formal AI policies in place, and even fewer enforcing them consistently.