TLDR: IBM leaders anticipate 2025 will be a pivotal year for AI, with a significant shift towards proactive “agentic AI” and Large Language Models (LLMs) driving IT optimization, automation, and new enterprise possibilities. This transformation will necessitate new governance frameworks, upskilling of employees, and a focus on secure, ethical deployment to maximize ROI and augment human capabilities.
IBM executives are forecasting a profound transformation in IT optimization and enterprise operations throughout 2025, largely driven by the emergence of proactive AI agents and advanced Large Language Models (LLMs). This shift marks a fundamental departure from traditional AI tools towards more autonomous and collaborative AI systems .
According to Ritika Gunnar, General Manager, Data & AI at IBM, “As agentic AI emerges as a predominant theme in 2025—marking a fundamental shift from traditional AI tools to proactive agents and teams of agents—so too will questions around accountability and control of these increasingly autonomous systems.” This perspective is echoed by a survey conducted by IBM and Morning Consult, which found that an overwhelming 99% of developers building AI applications for enterprise are already exploring or developing AI agents, signaling 2025 as “the year of the agent” .
The Rise of Agentic AI and its Capabilities:
An AI agent is defined as a software program capable of autonomously understanding, planning, and executing tasks, powered by LLMs and able to interface with various tools and systems . Unlike traditional AI assistants that require constant prompting, agents can generate plans and carry out tasks independently, particularly effective when focused on specialized tasks and collaborating with other agents on complex, multi-part requests .
Chris Hay, a Distinguished Engineer at IBM, highlights four key developments enabling this surge: better, faster, smaller models; Chain-of-Thought (COT) training; increased context windows; and enhanced function calling capabilities. These advancements mean current models are sufficiently robust to power the envisioned AI agents .
Impact on Workflows and Human Collaboration:
The integration of agentic AI is expected to streamline and alter jobs, drive optimization, and handle mundane tasks, freeing humans for more creative and higher-level pursuits . Jill Goldstein, Global Managing Partner, HR & Talent Transformation at IBM Consulting, notes that companies will need to reevaluate work processes and create new team structures where humans oversee groups of autonomous AI agents . The consensus among IBM experts is that AI agents will augment human workers rather than entirely replace them, with humans retaining final decision-making authority, especially for complex scenarios .
Challenges and Governance:
The rapid adoption of agentic AI also brings significant challenges, particularly around governance, security, and ethical deployment. Ritika Gunnar emphasizes the need for “greater attention to the guardrails, processes and tools for how we govern agents, in order to build trust for this powerful new frontier of AI capabilities.”
Nataraj Nagaratnam, CTO, IBM Cloud Security, warns about “shadow AI,” where employees might use public AI tools with sensitive information, necessitating strategic enterprise actions to gain visibility and control over AI usage . Tina Tarquinio, Chief Product Officer, IBM Z and LinuxONE, suggests keeping certain AI workloads on-premises, especially for high-volume transactional data, to enhance security, resiliency, and compliance .
Vyoma Gajjar, an AI Technical Solutions Architect, stresses the importance of rigorous stress-testing in sandbox environments, designing rollback actions, and ensuring audit logs for high-stakes industries . Maryam Ashoori, Director of Product Management, IBM watsonx.ai, points out the critical need for transparency and traceability of actions for every agent activity to prevent issues like accidental data leakage or deletion .
Automation, Open Source, and Multimodal AI:
The synergy between AI and automation is also a key prediction for 2025. Bill Lobig, VP, Product Management, IBM Automation, states, “Automation is needed to solve AI’s complexity,” enabling organizations to scale AI initiatives confidently and proactively detect and resolve IT issues . This pairing can also contribute to sustainability goals by optimizing resource management and reducing data center strain .
Bill Higgins, watsonx Platform Engineering and Open Innovation, IBM Research, anticipates open-source AI solutions will become a dominant force, offering friendlier cost structures, greater transparency, and support for multi-cloud architectures, helping enterprises scale beyond experimentation . Furthermore, Sriram Raghavan, VP, IBM Research for AI, highlights the growing capability of multimodal AI models to process and analyze complex documents with embedded rich content like images, tables, and charts, unlocking new possibilities for insights .
Strategic Implementation for ROI:
Also Read:
- Gartner Predicts 40% of Enterprise Applications to Integrate Task-Specific AI Agents by 2026
- Industry Surveys Reveal Surging Adoption of AI Agents and Evolving Data Science Landscape in 2025
Ultimately, IBM experts advise enterprises to develop a robust AI strategy focused on economic value and to get “agent-ready” by organizing proprietary data to power agentic workflows. Maryam Ashoori concludes, “Last year was the year of experimentation and exploration for enterprises. They need to scale that impact and maximize their ROI of generative AI. Agents are the ticket to making that happen.”
