AI Agents Represent a $450 Billion Market Opportunity Through 2028, Capgemini Research Finds

Capgemini's prediction that agentic AI has the potential to generate $450 billion in economic value by 2028 sits on a nascent foundation of trust that every AI company is attempting to strengthen and scale.

The problem is that most C-level execs aren't sure they can trust agentic AI or chatbots yet.

 Capgemini's latest report, Rise of agentic AI: How trust is the key to human‑AI collaboration, found that while 93% believe early adopters will gain a competitive advantage, only 27% trust fully autonomous agents with critical business processes. Company leaders clearly aren't ready to hand over the reins to agentic AI today. 

Not surprisingly, 61% of enterprises report rising employee anxiety about the impact of AI agents on their employment prospects, and over half believe AI agents will displace more jobs than they create.

The anxiety and fear that agentic AI is creating are fueling a swift rate of shadow AI app development and adoption. See VentureBeat's latest analysis of just how pervasive shadow AI is and how fast it is growing, as more professionals view it as a form of stealth insurance to keep their jobs.

Capgemini demystifies why AI agents aren't just more innovative chatbots

To fully appreciate the dynamics of how agentic AI has the potential to improve the productivity of enterprises adopting it, Capgemini went all-in on defining the differences between gen AI agents and chatbots. The researchers who wrote the report define ChatGPT, Microsoft Copilot, Google Gemini, and Le Chat by Mistral AI as gen AI assistants.

Traditional AI assistants are differentiated by being limited to performing specific tasks that are initiated exclusively by user prompts and pre-trained logic hard-wired into their architectures. What makes AI agents so unique, according to Capgemini's research team, is an AI agent's ability to independently, even taking initiative based on scheduling and logic workflows to make decisions, anticipate needs, and handle sophisticated workflows across multiple enterprise systems.

Capgemini chose to make it very clear in their report how they see each of the leading AI agents making contributions to organizational productivity today.

• Operator (OpenAI) autonomously performs repetitive tasks such as filling forms, booking reservations, and purchasing online.
• Devin independently manages coding tasks, debugging, and software deployment without continuous oversight.
• Manus handles diverse assignments, including personalized itinerary creation, research, and advanced document analysis.

Capgemini also clarifies the broader concept of agentic AI, which covers the entire ecosystem of technologies that empower agents. Agentic systems existed long before the current AI boom, but advancements in large language models and generative AI are now fueling their rapid enterprise adoption.

AI agents aren't smarter chatbots; instead, they represent a new class of independent decision-making automated entities driving measurable enterprise outcomes.

Capgemini quantifies AI agent capabilities, accelerating while inference costs continued to plummet

Capgemini's latest research confirms two powerful trends reshaping the economics of AI that are also paying out across the entire AI infrastructure landscape. Two of the most significant drivers are exponential improvements in agentic model capabilities and sharply declining inference costs.

As the graphic above illustrates, agentic AI's ability to autonomously handle complex, long-duration tasks has surged exponentially, especially since early 2023. At the same time, the cost of AI inference, or the computing required to deploy these models at scale, has dropped over 100x in just the past two years.

Every major player from NVIDIA to specialized ASIC vendors and the many open-source and proprietary AI and LLM providers are all now chasing the inference cost curve. The stakes are high because every one of these companies knows that if they can dominate the cost-performance curve of their area of the industry, it translates directly into market leadership and sustained enterprise adoption.

Last week, I took a look at how ASIC providers are gaining on NVIDIA. NVIDIA CEO Jensen Huang reinforced how differentiated and diverse the company he leads is, especially when it comes to being the leading AI infrastructure provider. NVIDIA has historically dominated these cost-improvement dynamics, but as inference costs keep plunging, its competitive position faces unprecedented challenges. I expect to see ASIC providers challenge NVIDIA more than ever in the coming years.

Where's the ROI for agentic AI coming from? Capgemini breaks it down

Capgemini's analysis provides clear answers on how AI agents drive financial impact, dividing it neatly into two sections: operational efficiency and top-line growth.

Operational efficiency improvements center on reduced cycle times, document automation, accelerated approvals, and streamlined IT workflows. These are table stakes for driving value to the bottom line of any business choosing to invest in agentic AI.

Capgemini is predicting revenue growth will be strongest in the following strategy areas, which they follow both as a research institute and a leading consultancy: accelerated drug discovery, new autonomous products, and the ability to offer 24/7 services and personalized customer experiences.

Capgemini finds enterprise adoption of AI agents surging 3.5 times faster than predicted

One of the most valuable aspects of Capgemini's commitment to research is how accurate and precise it strives to be when assessing new technology adoption. Their reports are among the best in the industry in this area. Because of their track record, their numbers on AI adoption seem to be the most reasonable and rational.

They're seeing AI agent adoption accelerating well beyond projections, rising sharply from just 4% to 14% in a single year. That's a 3.5x jump that tracks closely with the rapid expansion of generative AI.

The researchers who wrote the report share that early use cases of AI agents delivering measurable results across a spectrum of complex industry problem areas further validate the value they are providing. The report cites Ericsson uses AI agents to autonomously predict and resolve network disruptions, while Siemens deploys agent-based orchestrators to streamline automation on factory floors, driving faster innovation cycles.

Despite these big wins in two companies, world-renowned for their engineering and executive expertise, enterprise leaders continue to remain cautious. Capgemini explains that most deployments favor semi-autonomous AI agents operating within clearly defined guardrails and strict human oversight, underscoring a careful but committed path forward.

Customer service, IT, and sales lead AI agent adoption

High-volume tasks that have slight variation in the process workflows are prime candidates for being automated with agentic AI. The lack of variation also translates into a quicker learning curve for models to pick up any slight aberrations or exceptions in workflows, something that, from personal experience, I've seen can trip up a model you're trying to teach an application workflow or logic branching step.

One CEO told me in confidence that agentic AI is being given the work everyone was procrastinating about because it was so mundane and tedious to do. This includes electronic filing, analyzing and updating customer service reports, and tracking IT  operations exceptions. The CEO, who has a strong sales background, joked that agentic AI may finally achieve her goal of getting the majority of customer call reports into CRM while taming the nemesis of all sales teams: getting expense reports in before the end of the quarter.  

Her perspective aligns well with what Capgemini found. Customer service, IT operations, and sales are outpacing all other functions in AI agent adoption, according to the study.

The research shows that over the next three years, deployment will widen significantly in those core functional areas. Customer service adoption is predicted to climb to 45%, IT operations will grow to 40%, and sales will reach 39%.

Why AI agents are facing a crisis of trust in the enterprise

Capgemini also found that executives' trust in AI agents is eroding, with their growing fears around ethics, safety, and transparency. Given those factors, it is understandable why executive confidence in fully autonomous agents plummeted from 43% in 2024 to just 22% this year.

Four in ten executives believe AI agents' risks also outweigh their benefits. Top concerns include data privacy, cybersecurity vulnerabilities, hidden biases, and persistent opacity in AI decision-making.

Every organization is seeing the trust gap, and more boards and their senior leadership teams are surfacing concerns. More CISOs are getting called in to see if there is a way ot minimize risk and still get the productivity gains autonomous AI agents can deliver.

Bottom Line

Trust is at the center of whether Capgemini's prediction that agentic AI will generate $450 billion or not will come true. Adoption surged more than 3.5 times in one year, led by customer service, IT, and sales. But full autonomy remains rare. Executive confidence in fully autonomous AI dropped sharply from 43% to 22% this year, as concerns over bias, transparency, and privacy became real.

Getting trust right is going to be the most challenging aspect of integrating agentic AI into an enterprise. The economics are compelling and a strong motivator; it remains to be seen if the vendors driving sales will be able to overcome these challenges and grow the market.