IDC study: Organizations embrace the agentic future

Despite just half (50 percent) of organizations reporting having 10 or more agents in production in 2025 and healthy development pipelines via third-party applications, moving beyond the pilot stage remains a challenge. Less than seven percent are in full production with at least one use case—which demonstrates a clear and urgent need for deployment accelerators, as well as embedded agents that can more easily plug into existing stacks.

Organizations overwhelmingly want customization: Almost two thirds buy pre-built agents to customize, bemoaning the lack of industry-specific data training/terminology and vertical integration off the shelf. Software companies that ship agents with rapid customization options will rise to the top, especially if they can innovate fast enough to service the 90.9 percent expecting vertical industry and domain-specific agents, as well as extended memory agents able to accomplish longer tasks and goals.

As with the first wave of AI adoption, skills challenges, observability, integration issues, and cost concerns are hampering progress. To solve this, software companies and cloud providers (61 percent of custom agents are built using cloud-native APIs) must work together in the ecosystem to better help organizations navigate the issues. High value wins include easier, faster access to training at all levels, flexible deployment methods, simplified and well-supported paths for increasingly complex builds, and greater transparency on costs.

Organizations are pursuing sophisticated implementations, mixing custom-built and purchased agents (44 percent), multi-agent systems (37 percent), and agents from different providers in the same workflow (28 percent). To succeed, software companies will have to prioritize and perfect the specific technical capabilities these same organizations cite as top challenges, including latency/responsiveness, output accuracy, and API issues.

More than two thirds (67 percent) of organizations believe users need more skills training to increase agentic AI adoption, with lack of skilled personnel (55 percent) cited as the top implementation challenge. Talent and technology must connect. Software companies will benefit from bundling role-based training and in-product guidance with agents plus certifications. Added to this, success metrics should tie upskilling directly to adoption and business outcomes.