How Ai Agents Can Generate $450 Billion By 2028 - And What Stands In The Way

Philip Thurston/Getty Images

Agentic AI is 1 of nan fastest-emerging technologies successful business, pinch nan imaginable to make $450 cardinal successful economical worth done gross uplift and costs savings crossed surveyed countries by 2028, according to the Rise of agentic AI: How spot is nan cardinal to human-AI collaboration. The caller study from Capgemini Research Institute reveals insights from a study of 1,500 elder executives crossed 14 countries connected nan emergence of AI agents arsenic a transformative unit successful business. 

Also: My 8 ChatGPT Agent tests produced only 1 near-perfect consequence - and a batch of replacement facts

Key findings from nan Capgemini report

• A $450 cardinal opportunity: By 2028, AI agents could dress up to $450 cardinal successful economical worth done gross maturation and costs savings crossed surveyed markets.
• Adoption is accelerating, but maturity remains low: Only 2% of organizations person deployed AI agents astatine scale, 12% astatine partial scale, 23% person launched pilots, and 61% are exploring deployment. Fifteen percent of business processes are expected to scope semi- aliases afloat autonomy successful nan adjacent 12 months. However, less than 1 successful 5 organizations study precocious maturity pinch respect to nan information and exertion infrastructure needed to instrumentality agentic AI.
• Trust successful afloat autonomous AI agents is declining: It dropped from 43% to 27% successful 1 year. Ethical concerns, deficiency of transparency, and constricted knowing of agentic capabilities are cardinal barriers. Ethical concerns astir AI, specified arsenic information privacy, algorithmic bias, and nan "AI achromatic box," are prevalent, but fewer organizations enactment decisively to mitigate them.
• Human-AI collaboration is nan future: By 2028, 38% of organizations will person AI agents arsenic squad members wrong quality teams. Blended teams -- wherever humans and AI agents collaborate -- will go nan norm, driving productivity and innovation. This intends that successful 3 years, organizations expect to person AI agents arsenic members wrong human-supervised teams; AI agents must beryllium seen not arsenic tools, but arsenic portion of nan team.

What are AI agents and agentic AI? 

Here is Capgemini's definition: "AI agents are programs/platforms/software that are connected to nan business situation pinch a defined boundary, make decisions autonomously, and enactment to execute circumstantial goals pinch aliases without quality intervention. With nan latest advances successful reasoning AI models, AI agents are capable to break down tasks, 'reason' done imaginable pathways to find solutions to nan fixed problem, effort those solutions, and coming successful outcomes."

Capgemini notes that nan capabilities of AI agents are expanding fast, and nan costs to run and create AI agents is declining. AI agents are 1 of nan apical exertion trends for 2025.

My 10 cardinal takeaways from Capgemini's investigation  

1. AI agents' effect connected nan economy

AI agents could make astir $450 cardinal successful full economical worth successful nan 14 countries surveyed by 2028. Organizations pinch scaled implementation are projected to make astir $382 cardinal (2.5% of yearly revenue) connected mean complete nan adjacent 3 years, while we expect others to make astir $76 cardinal (0.5% of yearly revenue). This applies to an mean statement pinch $15 cardinal successful yearly revenues. Capgemini expects that surveyed organizations will collectively execute $19 cardinal successful gains complete nan adjacent 12 months, pinch this fig projected to summation to $92 cardinal by nan 3rd year.
Other predictions extracurricular of Capgemini include:

Also: Is HR fresh for AI?

Goldman Sachs predicts that Gen AI will thrust a 6.1% emergence successful US GDP complete nan adjacent decade -- by 2028, this translates to astir $540 cardinal successful nan US.

IDC forecasts that AI technologies wide will power 3.5% of world GDP by 2030 -- by 2028, this implies an effect of astir $1.9 trillion globally.

MIT investigation estimates that a operation of AI capabilities could automate astir one-fifth (slightly complete 20%) of value-added tasks.

2. Business assurance successful AI agents

A important 93% expect that organizations that person successfully scaled nan implementation of AI agents wrong nan adjacent 12 months will execute a competitory advantage. 

Also: Business leaders proceed to push workers toward regular usage of AI

3. Adoption of AI agents is connected nan rise

Currently, 23% of organizations person initiated AI supplier aviator projects, while 14% person progressed to partial aliases full-scale implementation. About 30% are exploring AI agents, and different 31% are preparing for experimentation aliases deployment wrong nan adjacent six to 12 months. In fact, nan gait of AI supplier take mirrors nan accelerated trajectory seen pinch generative AI.

4. Businesses connected on their measurement to becoming autonomous 

Many organizations that declare to beryllium implementing AI agents are deploying solutions pinch constricted autonomy. As galore arsenic 85% of business processes are expected to beryllium astatine debased levels of autonomy successful nan adjacent 12 months. Industries noted successful nan Capgemini study see automotive, financial services, life sciences, telecom, and unit -- pinch circumstantial institution examples and usage cases.

Also: Your adjacent job? Managing a fleet of AI agents

The study noted that 62% of companies for illustration to partner pinch solution providers, specified arsenic Salesforce and strategy integrators, to instrumentality aliases tailor already disposable AI agents. The fresh readiness of in-built agents, pre-existing integrations pinch bequest systems, and unit fluency successful utilizing these devices apt lend to nan preference.

5. AI supplier early roadmaps are evolving

Capgemini recovered that 16% of organizations person developed a strategy and roadmap to instrumentality agentic AI. Research recovered that 39% of organizations don't person a strategy, but person aggregate initiatives crossed functions to create innovative solutions that tin beryllium scaled up.

6. AI activity is cardinal to accelerating AI supplier roadmaps

The request for dedicated activity to oversee AI supplier initiatives is besides a priority, pinch 26% appointing caller leaders specifically for AI agents and 59% delegating this work to existing AI aliases gen AI leadership.

7. Preferred agentic AI pricing models vary

Over half of organizations for illustration consumption-based pricing for AI models wrong AI agents. Consumption-based (55%), platform-based (43%), and license-based(37%) models are nan preferred pricing models among organizations for agentic AI solutions.

8. Customer-facing business functions astir apt to adopt AI agents

Customer services and support, IT, and income are nan functions wherever astir executives foretell that AI agents will beryllium actively performing astatine slightest 1 process aliases sub-process regular wrong nan adjacent 12 months. These functions often impact precocious volumes of interaction, require responsiveness complete precision, and dangle connected contextual, conversational engagement.

Also: The decease of spreadsheets: 6 reasons why AI will soon beryllium nan ascendant business reporting tool

9. Fully autonomous business capabilities are a longer journey

Digital labour will grip 1 process by 2028, arsenic 58% of business functions are apt to person AI agents handling astatine slightest 1 process aliases sub-process regular wrong 3 years. Over nan adjacent 12 months, we expect AI agents pinch Level 3 autonomy aliases higher to negociate 15% of processes and subprocesses successful each business function. This will emergence to 25% wrong nan adjacent 1 to 3 years. Fully autonomous AI agents (Level 5) are expected to grip astir 4% of business processes wrong 3 years. Capgemini noted that successful nan adjacent 12 months, it expects AI agents to make 6% of day-to-day decisions, expanding to 8% successful 1 to 3 years. And 25% of processes wrong a business usability are expected to beryllium handled by AI agents pinch Level 3 aliases higher autonomy by 2028.

10. Trust successful AI agents is cardinal to accelerating adoption

Almost half (47%) of organizations successful nan implementation shape study supra mean level of spot successful AI agents, compared pinch 37% that are still successful nan exploration phase. This beardown relationship betwixt spot and implementation confirms that spot plays a important domiciled successful nan take of AI agents. Only 22% of executives spot afloat autonomous AI agents for endeavor applications, down from 43% successful 2024. The spread is widening successful nan incorrect direction. 

Also: Beyond nan AI spot gap: A leader's playbook

The diminution successful spot is not constricted to AI agents; it extends to AI and gen AI arsenic well. Capgemini noted that half of organizations person insufficient knowledge of AI agents' capabilities. Employee skills and AI infrastructure maturity adhd to nan spot deficiency. Over 4 successful 5 (82%) organizations study low-to-medium maturity crossed dimensions specified arsenic computing, integration, orchestration, fine-tuning, and cybersecurity.

The Capgemini study concludes pinch this powerful reminder: 

"The winners successful this adjacent activity of AI will not beryllium those who simply deploy much AI tools. Rather, they will beryllium those who rethink their business, reimagine workflows, reskill their workforces, restructure their organizations, and embed ethical safeguards from nan outset." 

I judge AI agents whitethorn correspond nan biggest technological effect connected businesses ever. This is not astir caller technology. This is astir a caller world. The astir successful and resilient businesses of nan early will beryllium autonomous. Autonomous businesses will person a hybrid workforce -- humans and integer labor, leveraging AI agents to augment and upgrade quality capabilities. AI agents will beryllium utilized for some cognitive transportation and cognitive upgrade opportunities. Autonomous businesses will besides position themselves to champion return advantage of the caller stability-performance business operating models. 

Also: Why smart businesses usage AI to offload tasks and supercharge their teams

To study much astir the Rise of agentic AI: How spot is nan cardinal to human-AI collaboration report from the Capgemini Research Institute, you tin sojourn here.

Want much stories astir AI? Sign up for Innovation, our play newsletter.