AIToday
Large Language ModelsAI Business & IndustryAINOWPublished: Oct 3, 2026, 01:00 JST

Gartner: 40%超 of AI agent projects to be canceled before 2027年末

Gartner: 40%超 of AI agent projects to be canceled before 2027年末

3 Key Points

  1. What happened

    Gartner predicts over 40% of agentic AI projects will be canceled by the end of 2027. Yano Research Institute found only 3.3% of Japanese generative AI users are already using AI agents.

  2. Why it matters

    Most Japanese companies remain in the research or consideration stage, according to Yano's survey. The challenges differ by phase — purpose, data, and talent before adoption; integration, accuracy, and penetration after.

  3. What to watch

    Okta's survey found 34% of organizations apply security controls to AI equivalent to human access. Whether companies formalize agent governance before scaling will likely determine which projects survive.

WHO IT HITSJapanese corporate teams tasked with evaluating AI agents will likely find the most value in phase-specific planning data, particularly IT managers and DX promotion staff who must justify investments to executives using Gartner's cancellation forecast.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, in your inbox every morning.

Context & Analysis

The survey landscape reveals a sharp gap between interest and action. Yano Research Institute's 2025 survey found that while over 60% of Japanese companies express interest in AI agents, only 3.3% have moved to actual use, with 13.5% in consideration and 49.3% still gathering information. This gap suggests the barriers to adoption are substantial and not easily resolved by enthusiasm alone.

Grasys's survey (580 respondents at Google Cloud Next Tokyo '26) and JIPDEC's 2026 enterprise IT survey both point to a structural pattern: the obstacles companies face shift as they move from planning to operations. Before adoption, companies struggle to articulate what they want AI agents to do; after adoption, the difficulty becomes integrating agents with existing systems and keeping usage from concentrating among a few employees. Several Japanese companies have begun to address these challenges. Resonac Holdings standardized agents created on the front lines and rolled them out company-wide, reaching a 95% utilization rate about six months after introduction. Panasonic Connect tracked monthly unique user rates, reaching 61% for its ConnectAI assistant. GMO Internet Group combined tool subsidies with talent development, achieving a 71.4% business utilization rate.

Whether such success stories become the norm or the exception likely hinges on how companies handle governance. Okta's 2026 survey across seven countries found 58% of executives experienced AI-related security incidents in the past 12 months, yet only 34% of organizations applied the same security controls to AI as to human access. If that governance gap persists while agent deployments scale, Gartner's prediction of over 40% project cancellations by the end of 2027 may prove conservative — though the companies that build governance and measurement frameworks early could pull ahead of peers still stuck in the consideration stage.

FAQ
Why do AI agent projects get canceled?
Gartner cites rising costs, unclear business value, and inadequate risk management as the reasons agentic AI projects are canceled. AI agents also use usage-based billing on top of license fees, making costs hard to estimate upfront.
What's the difference between generative AI and AI agent adoption challenges?
AI agents execute business tasks, so integration and governance challenges are larger than with generative AI. Generative AI challenges centered on information leaks, wrong answers, and employee usage rates; AI agents add system integration and access permission management.
What is agent washing?
Gartner points out that vendors are engaging in 'agent washing,' labeling existing chatbots or RPA as AI agents. Companies should verify whether vendor offerings actually match their objectives rather than accepting vendor claims at face value.

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.

Questions and answers are published on this page.

Related Articles

Next articleGwen Davis: Direct AI agents, don't just use them