TLDR : The large-scale deployment of AI agents, promising but not yet widely implemented, depends more on business trust than technology, according to the Capgemini Research Institute. The main obstacles remain ethical concerns, immature infrastructure, and insufficient data readiness, despite a considerable potential for economic gains.
Table of contents
While 2025 was touted as the year of maturity for AI agents, their deployment will depend less on technology and more on trust, according to the latest report from the Capgemini Research Institute. Despite a potential economic impact of $450 billion by 2028, only 2% of companies have effectively deployed them on a large scale.
For its study "Rise of agentic AI: How trust is the key to human-AI collaboration", the research institute of the IT services company surveyed 1,500 business leaders generating over a billion dollars in annual revenue, spread across 14 countries, all of which had begun exploring agentic AI.
Enthusiasm Curbed by a Significant Decline in Trust
The initial excitement surrounding agentic AI was based on enticing promises: freeing up human time, accelerating decision-making, and restructuring entire support or operational functions. However, in just one year, trust in fully autonomous AI agents dropped from 43% to only 27%. Ethical concerns (privacy, bias, lack of explainability) hinder adoption: 2 out of 5 leaders believe the risks outweigh the benefits.
Nevertheless, companies that have already implemented AI agents show a trust level 10 points higher than the average.
Humans Remain Essential
The emerging model is no longer one of AI replacing humans, but rather hybrid pairings where agents and employees form a distributed team. Nearly 60% of surveyed organizations plan to establish this model within the next 12 months.
As noted by Niraj Parihar, CEO, Insights and Data Global Business Line, Capgemini: "AI agents are team members, not replacements."
The expected benefits of this collaboration are tangible: respondents anticipate not only a 65% increase in human engagement in high-value tasks but also a 53% gain in creativity and a 49% improvement in employee satisfaction.
Organizational Barriers Still Present
Despite the enthusiasm of 93% of surveyed leaders, only 2% of companies have deployed AI agents on a large scale. The main barriers are an immature infrastructure for 80% of them and insufficient data readiness (nearly 4 out of 5 companies).
In this context, only 15% of functions are expected to integrate AI agents at a semi-autonomous or autonomous level by 2026. This rate could reach 25% by 2028, according to Capgemini.
AI agents are expected to first find their place in areas such as customer service, IT, and commercial functions, where their automation capabilities offer quick gains. Their adoption should then gradually expand over the next three years to more complex sectors like operations, R&D, and marketing.
To fully harness their potential, the report recommends moving beyond the trend and rethinking processes, reinventing business models, transforming organizational structures, and ultimately finding the right balance between agent autonomy and human involvement.
Franck Greverie, Chief Portfolio and Technology Officer at Capgemini, head of Global Business Lines and member of the Group's Executive Committee, concludes:
"To succeed, businesses must stay focused on outcomes, rethinking their processes with an 'AI first' approach. The success of this transformation lies in the need to build trust in AI by ensuring it is developed responsibly, integrating ethics and security by design. This also involves reorganizing companies to foster effective collaboration between humans and AI, and creating the right conditions for these systems to enhance human judgment and improve economic performance."