The rise of AI agents: turning generative AI into autonomous problem‑solvers

Few technologies have vaulted from research to ubiquitous conversation as quickly as generative AI. Less than three years after ChatGPT reached 100 million users in just two months, the technology world has begun talking about agents – software systems that don’t just generate text but can autonomously plan, act and learn. Early experiments show promise: a consumer goods company using agents to draft blog posts reduced costs by 95 % and delivered content 50 times faster[1]. A global bank that deployed AI virtual agents cut customer‑interaction costs ten‑fold[1]. Even so, many observers are wary. Gartner predicts more than 40 % of agentic AI projects will be cancelled by 2027 because of soaring costs and unclear value[2], and trust in fully autonomous systems is falling[3]. This blog explores the landscape behind these headlines: what AI agents are, why the market is booming, where early value is emerging and what hurdles still lie ahead.

From models to agents: what makes an AI agent?

An AI agent is a software system built on top of large language models (LLMs) that can understand goals, plan actions, invoke external tools and continuously learn. Unlike chatbots that respond to individual prompts, agents operate with a degree of autonomy: they break down tasks, call APIs or software to complete subtasks, and adapt their behaviour based on feedback. Leading AI providers like OpenAI, Google, Microsoft, Anthropic and Meta now offer agentic platforms that wrap LLMs with reasoning, memory and orchestration capabilities. Specialist providers have also emerged, from sales‑oriented agents (HubSpot’s Breeze) to software‑engineering agents (OpenAI’s A‑SWE) and research assistants like Deep Research[4]. These systems integrate technologies such as planning algorithms, vector‑search memory, tool execution and reinforcement learning to achieve complex tasks autonomously.

A booming market: growth forecasts and investment

Market research firms agree that the market for AI agents is surging from a tiny base. In 2024 the global AI agents market was valued at about USD 5.4 billion[5]. Precedence Research projects the market to grow to USD 7.92 billion in 2025 and USD 236 billion by 2034, implying a compound annual growth rate (CAGR) of ~45 %[6]. Other analysts report similar trajectories: Markets & Markets expects growth from USD 5.1 billion in 2024 to USD 47.1 billion by 2030 (CAGR 44.8 %)[7]; Grand View Research cites a CAGR of 45.8 % between 2025 and 2030[5]; and data provider Precedence Research notes that North America holds roughly 41 % of the market while Asia‑Pacific is the fastest‑growing region[8].

Investors are betting on this growth. The venture capital market poured around $2 billion into agentic AI start‑ups by mid‑2025[9], and a surge of open‑source frameworks (e.g., LangGraph, AutoGen and CrewAI) have attracted tens of thousands of developers and millions of monthly downloads[10]. These frameworks allow companies to experiment cheaply with autonomous workflows before committing to enterprise‑grade implementations.

Adoption across industries: early pilots to production

Enterprise appetite

A PwC survey of senior executives in May 2025 found that 88 % planned to increase AI budgets, 79 % reported they were already adopting AI agents, and 66 % of adopters said agents delivered measurable value[11]. Yet only 35 % had adopted agents broadly across the organization, and just 17 % fully integrated them[12]. Another poll by EY covering 504 US tech leaders reported that 48 % were integrating or fully deploying agentic AI, while 92 % planned to increase AI spending[13]. UiPath’s 2025 Agentic AI report surveyed 252 US IT executives and found 93 % were extremely or very interested in the technology, 45 % were ready to invest this year, and more than 30 % planned to invest within six months[14].

LangChain’s State of AI Agents 2024 survey also showed that adoption is no longer confined to large tech firms: 51 % of respondents already run agents in production and 78 % plan to implement more soon[15]. Adoption is particularly high among mid‑sized companies (100–2,000 employees) and 90 % of non‑technology companies plan to deploy agents, nearly matching technology firms[15]. Top use cases include research and summarization (58 %), personal productivity assistants (53.5 %), customer service (45.8 %) and code generation (35.5 %)[16].

Early results and ROI

Returns vary by use case but early pilots suggest meaningful efficiency gains. Verizon reported a 40 % increase in sales after deploying a Google AI sales assistant to support 28,000 customer‑service representatives[17]. ServiceNow’s integration of AI agents cut the time required to handle complex service cases by 52 %[18]. A survey by PagerDuty found 62 % of companies expect 100 % or greater return on investment from agentic AI deployments[19]. BCG’s consulting clients have achieved cost reductions up to 95 % and 40× speed improvements for content creation[1]; other pilots have cut cycle times in lead generation by 25 % and boosted productivity in IT modernization projects by 40 %[1]. McKinsey estimates that across marketing and sales, generative AI (a precursor to agentic systems) can deliver 3 %‑15 % revenue uplift and 10 %‑20 % sales ROI improvement[20].

Sector‑specific trends

The distribution of use cases is uneven. Advisory firm ISG notes that over half of agentic AI use cases today occur in IT functions, with marketing, sales and finance each contributing only about 10 %[21]. Among providers surveyed, 70 % of agentic projects are concentrated in three industries – banking/financial services, retail and manufacturing[22]. In manufacturing, AI‑driven predictive maintenance has reduced downtime by 40 %, and in healthcare, AI scribes like Abridge have improved patient comprehension by 40 %[23]. Warmly’s statistics collate industry‑specific adoption: 90 % of hospitals are expected to adopt agents by 2025[24]; 69 % of retailers report revenue growth from personalized AI‑driven shopping experiences[25]; and 88 % of marketers leverage AI agents to accelerate content creation and decision making[26].

Vertical agents: beyond generic chatbots

Traditional generative AI tools are horizontal: they can write code or summarize text but are not optimized for specific industries. Investment managers at Wellington argue that vertical AI agents – systems tailored to sectors like healthcare, government or small business – could unlock a larger total addressable market than the entire US$315 billion enterprise software industry[27]. These agents process unstructured data (e.g., clinicians’ notes, voice recordings) and integrate with legacy systems to execute specialized workflows[28]. In healthcare, 80 % of medical data is unstructured and the WHO predicts a shortfall of 18 million healthcare workers by 2030[29]. AI scribes that automate documentation and charting can free clinicians’ time and improve accuracy; the University of Chicago Medicine reported a 40 % increase in patient comprehension after using Abridge’s AI scribe[23]. In government, agentic systems can handle routine permit processing and voter information; there are over 90,000 local government units in the US, and AI agents trained on municipal data could help manage the US$3.9 trillion annual state and local spending[30]. For small‑ and medium‑sized businesses, which number 400–500 million worldwide, agents can automate back‑office tasks like invoicing, customer service and inventory, enabling SMBs to compete at enterprise‑level efficiency[31].

Why is the hype high? Underlying drivers and enablers

Workers face information overload

Knowledge workers today spend hours navigating email, documents and meetings; an average employee receives over 120 emails per day (roughly three hours of reading)[32]. AI agents promise to triage and respond to routine messages, freeing workers for creative tasks. According to Gartner, by 2028 agents will autonomously make at least 15 % of day‑to‑day work decisions and be embedded in 33 % of enterprise software applications[33], up from less than 1 % in 2024. This shift could reduce cognitive load and increase productivity across white‑collar roles.

Cheaper compute and open‑source tooling

The cost of training large models is falling fast; Chinese start‑up DeepSeek recently claimed a 94 % decline in machine‑intelligence costs[34]. Meanwhile, open‑source frameworks such as LangGraph (14 k GitHub stars, 4.2 million monthly downloads)[10], AutoGen (45 k stars) and CrewAI (32 k stars)[35] make it easier for developers to build multi‑step, multi‑agent workflows. Enterprises can pilot agents at low cost and gradually scale successful prototypes.

Strategic imperatives and competitive pressure

Agentic AI promises to break what McKinsey calls the “generative AI paradox”: while eight in ten companies have deployed generative AI, most vertical use cases remain stuck in pilot mode and earnings gains are limited[36]. By automating end‑to‑end workflows rather than merely generating content, agents offer a path to unlock tangible business value. PwC found that 75 % of executives believe AI agents will reshape the workplace more than the internet[11], and 46 % worry their organizations will fall behind if they don’t adopt agentic AI[12]. Companies that successfully deploy agents could gain a significant first‑mover advantage.

Barriers and risks: hype meets reality

The enthusiasm comes with caveats. Reliability and error propagation remain central concerns; chaining multiple AI calls compounds hallucinations and can lead to unpredictable behaviour[37]. Safety and value alignment are complex because autonomous agents may pursue goals in unanticipated ways, requiring careful human oversight and “human‑in‑the‑loop” design[38]. Security is a pressing risk; prompt‑injection attacks or malicious tool calls could cause real‑world harm, prompting high‑profile incidents[39]. Scalability and cost also matter: computing resources for advanced agents remain expensive, and Gartner warns that more than 40 % of agentic AI projects will be cancelled by 2027 due to escalating costs and unclear value[2]. Integration is another hurdle: nearly 95 % of IT leaders report challenges integrating agents with existing systems[40].

Trust and governance are equally important. Capgemini’s Rise of Agentic AI report found that while 93 % of leaders believe scaling AI agents will provide a competitive edge, only 27 % of organizations trust fully autonomous agents[3]. Another 70 % of executives emphasised upskilling and 68 % plan to hire AI talent to manage the transition[13]. Regulatory uncertainty looms; compliance concerns climbed from 28 % to 38 % through 2024[41], and global governments are drafting rules on safety, data privacy and accountability.

Looking ahead: from pilots to pervasive agents

Despite the challenges, most analysts expect agentic AI to evolve from experimental pilots to pervasive enterprise infrastructure over the next decade. Deloitte predicts that 25 % of companies using generative AI will launch agentic AI pilots in 2025, rising to 50 % by 2027[42]. ISG anticipates that more than half of functional AI use cases will remain in IT while industries like financial services, retail and manufacturing continue to lead early adoption[43]. Gartner expects agentic AI components to be embedded in one‑third of software applications by 2028[33]. Meanwhile, venture funding and open‑source innovation are accelerating, and vertical agents tailored to healthcare, government and SMBs may unlock new markets beyond traditional enterprise software[27].

For organizations contemplating agentic AI, the message is clear: start small but think big. Experiment with clearly defined tasks, embed human oversight and robust guardrails, and invest in data quality and infrastructure. Train employees to work alongside agents and build transparent governance processes. Agents won’t replace humans; they will augment them, automating mundane work and amplifying human creativity. In the words of BCG’s research, AI agents are poised to become “teammates” rather than tools[1]. Harnessed responsibly, they could transform how we work—and open up markets that dwarf even the biggest technology revolutions to date.

[1] AI Agents: What They Are and Their Business Impact | BCG

https://www.bcg.com/capabilities/artificial-intelligence/ai-agents

[2] [33] Over 40% of agentic AI projects will be scrapped by 2027, Gartner says | Reuters

https://www.reuters.com/business/over-40-agentic-ai-projects-will-be-scrapped-by-2027-gartner-says-2025-06-25/

[3] Final-Web-Version-Report-AI-Agents.pdf

https://www.capgemini.com/wp-content/uploads/2025/07/Final-Web-Version-Report-AI-Agents.pdf

[4] [38] [39] Developments in AI Agents: Q1 2025 Landscape Analysis — The Science of Machine Learning & AI

https://www.ml-science.com/blog/2025/4/17/developments-in-ai-agents-q1-2025-landscape-analysis

[5] [10] [35] The Best AI Agents in 2025: Tools, Frameworks, and Platforms Compared | DataCamp

https://www.datacamp.com/blog/best-ai-agents

[6] [8] 150+ AI Agent Statistics [July 2025]

https://masterofcode.com/blog/ai-agent-statistics

[7] [37] Demystifying AI Agents in 2025: Separating Hype From Reality and Navigating Market Outlook | Alvarez & Marsal | Management Consulting | Professional Services

https://www.alvarezandmarsal.com/thought-leadership/demystifying-ai-agents-in-2025-separating-hype-from-reality-and-navigating-market-outlook

[9] Autonomous generative AI agents | Deloitte Insights

https://www.deloitte.com/us/en/insights/industry/technology/technology-media-and-telecom-predictions/2025/autonomous-generative-ai-agents-still-under-development.html

[11] AI agent survey: PwC

https://www.pwc.com/us/en/tech-effect/ai-analytics/ai-agent-survey.html

[12] 88% of US firms to increase AI budgets amid agentic AI adoption

https://www.techmonitor.ai/digital-economy/ai-and-automation/88-of-us-firms-increase-ai-budgets-agentic-ai-adoption

[13] Nearly half of tech firms now implementing agentic AI, survey finds

https://www.techmonitor.ai/digital-economy/ai-and-automation/agentic-ai-survey-ey-us

[14] UiPath Agentic AI Research Report 2025 | UiPath

https://www.uipath.com/resources/automation-analyst-reports/agentic-ai-research-report

[15] [16] LangChain State of AI Agents Report