AI agents to power 25% of enterprises by 2025, Deloitte predicts

The rapid evolution of enterprise AI adoption is reaching a critical turning point as businesses prepare to integrate more sophisticated autonomous AI systems into their operations.

Key forecast: Deloitte’s Global 2025 Predictions Report projects that 25% of enterprises using generative AI will deploy AI agents by 2025, with that number expected to double to 50% by 2027.

• Agentic AI, defined as autonomous software solutions capable of completing complex tasks with minimal human oversight, represents the next frontier in enterprise AI adoption
• This growth will be driven by both startups and established companies seeking new revenue opportunities
• The technology is expected to move beyond pilot programs and proof of concepts in certain markets by 2025

Technical capabilities: Agentic AI systems possess distinct characteristics that set them apart from traditional AI applications and chatbots.

• These systems are built on foundation models like Large Language Models (LLMs), enabling complex reasoning and adaptability
• They can operate autonomously, process multiple types of data (text, audio, video), and interact with various tools and systems
• Unlike standard LLMs, agentic AI can access both short-term and long-term memory to maintain context and learn from experience
• The technology can orchestrate multiple systems and collaborate with other AI agents

Market trends and adoption: The landscape of AI usage is showing significant shifts across different demographics and sectors.

• Women’s usage of generative AI is expected to match or exceed men’s by 2025, up from just half the usage rate in 2023
• By 2025, over 30% of shipped smartphones could feature generative AI capabilities
• Approximately 50% of laptops are expected to include local generative AI processing capabilities
• Data center electricity consumption is projected to double to 4% globally by 2030 due to AI-related power demands

Implementation challenges: Organizations face several hurdles in adopting these advanced AI technologies.

• Despite widespread recognition of AI’s potential, only 11% of CIOs report full implementation of the technology
• Security concerns and data infrastructure limitations remain primary obstacles
• CIOs are increasingly expected to serve dual roles as chief AI officers
• Consumer preferences show that one-third would choose AI agents for faster service, but transparency about AI interactions remains important

Strategic implications: The emergence of agentic AI represents a fundamental shift in how organizations will approach automation and decision-making processes, though success will likely depend on careful consideration of infrastructure requirements, security measures, and user acceptance.