Firms eye $450bn agentic AI prize, but trust is tanking

Companies are eyeing a projected $450bn (£336bn) economic windfall from agentic AI by 2028, but few are adequately equipped to get there.

New research from Capgemini found that the overwhelming majority of business leaders – 93 per cent – believe that scaling AI agents in the next 12 months will deliver a competitive ege.

Yet actual deployment remains limited. Just two per cent of firms have reached scaled adoption, and only 14 per cent have even begun implementing agents in live settings.

Most are still stuck in the planning or pilot phase.

Agentic AI – a new generation of systems capable of operating with little to no human oversight – promises cost savings and revenue uplift.

But, Capgemini warns that the reality is more cautious due to the trust factor.

Belief in fully autonomous AI agents has fallen sharply, from 43 per cent to 27 per cent over the past year, with ethical concerns and privacy fears contributing,

Just 40 per cent of organisations say they trust agents to operate independently.

Trust in agentic AI scale-up

Yet the report shows trust tends to increase as firms move from theory to practice.

Among firms actively implementing AI agents, 47 per cent report above-average trust levels, compared with 37 per cent of those still exploring.

Rather than aiming for full autonomy, many organisations are pivoting toward human-agent collaboration models.

Over the next 12 months, more than 60 per cent of companies plan to integrate AI agents into teams as copilots or assistants.

This approach is projected to unlock a 65 per cent increase in time spent on high value tasks, but raises questions on job stability in an increasingly tough market.

Lack of readiness

Despite the potential, readiness remains a drag on progress, with four in five firms lacking the infrastructure to scale, and fewer than one in give considering their data quality to be at the required level.

Meanwhile, concerns around algorithmic bias, privacy and the ‘black box’ nature of many agent decisions continue to stall uptake.

“The economic promise of AI agents is clear, but realising it means redesigning workflows and reshaping roles – not just bolting new tech onto legacy systems”, said Franck Greverie, Capgemini’s chief technology officer. “Trust, transparency and human-AI chemistry are now critical to unlocking value”.

Capgemini has estimated a $450bn opportunity, including both cost savings and revenue growth.

Scaled adopters could generate $382m in value on average over three years, which is five more than those still experimenting.

For now, most agents remain in support roles – with only 15 per cent of businesses processes currently operating at semi to fully autonomous levels.

That figure is expected to rise to 25 per cent by 2028, as organisations move beyond copilots toward more independent systems.