A survey of 625 IT professionals with cloud computing expertise finds that well over two thirds (69 percent) plan to source agentic artificial intelligence (AI) capabilities through IT or consulting service providers—second only to agentic AI platform vendors, which came in slightly higher at 71 percent.
Conducted by theCUBE Research on behalf of SOUTHWORKS, a provider of software engineering services, the survey reveals that 71 percent of respondents rely on some type of agentic AI capability from a platform vendor, while 59 percent work with a software‑as‑a‑service (SaaS) provider. More than half (51 percent) still depend primarily on public AI tools for implementation. Nearly half (47 percent) are working with open‑source frameworks and libraries, compared to 32 percent who plan to primarily build agentic AI capabilities in‑house.
The overall finding is clear: for now, most organizations prefer to buy AI agent technology and lean on third‑party expertise rather than build and maintain AI agents themselves.
How organizations are using AI agents today
AI agents today most commonly automate repetitive tasks (73 percent), support worker decision‑making (68 percent), and diagnose and solve business problems (66 percent).
Looking ahead, survey respondents plan to invest heavily in several advanced agentic AI capabilities, including:
- AI reasoning tools and platforms that can plan, optimize, and justify outcomes (71 percent)
- AI assistants that help execute tasks (71 percent)
- Autonomous, goal‑driven AI agents that take action (58 percent)
- Agentic workflows orchestrated by AI (52 percent)
- Multi‑agent collaborative systems (48 percent)
Why organizations still need third-party expertise
Regardless of the use case, many organizations are turning to IT service providers and consultants to better understand which processes are best suited for a probabilistic technology—one that may not produce the exact same answer or behavior every time.
Most business processes are deterministic and require strict repeatability, which makes the nuances of prompt engineering and agentic system design especially important. Organizations also want guidance on which agentic AI use cases deliver true competitive advantage versus those that are becoming basic table stakes for staying competitive.
The next phase: From isolated agents to connected workflows
Today, most AI agents primarily automate single tasks. Less than one third (30 percent) of organizations plan to deploy AI agents built on a common framework, while only 22 percent have already implemented one. Another 29 percent have limited agentic AI to isolated departmental use cases, and 24 percent report siloed deployments across multiple business units with no standardization.
These findings point to a major upcoming opportunity for IT service providers: helping organizations design, connect, and orchestrate workflows that span multiple AI agents.
It may take time before AI agents are deeply embedded across business processes, but the shift is becoming less a question of if and more a matter of when. For IT service providers, the challenge—and opportunity—is to help organizations deploy agentic AI safely, effectively, and where it delivers meaningful business value rather than unnecessary complexity.
Photo: Alfa Photo / Shutterstock
This post originally appeared on Smarter MSP.