{"id":66891,"date":"2026-04-21T09:08:50","date_gmt":"2026-04-21T08:08:50","guid":{"rendered":"https:\/\/sovanta.com\/?p=66891"},"modified":"2026-04-21T09:08:51","modified_gmt":"2026-04-21T08:08:51","slug":"why-enterprise-ai-agent-adoption-is-still-slow","status":"publish","type":"post","link":"https:\/\/sovanta.com\/en\/why-enterprise-ai-agent-adoption-is-still-slow\/","title":{"rendered":"Why Enterprise AI Agent Adoption is Still Slow"},"content":{"rendered":"\n

In this blog post series, Niklas Fr\u00fchauf<\/a>, Senior Data Scientist, explores whether AI agents already deliver real value for businesses. In part one<\/a>, he outlined the core concepts and different types of agents. In part two<\/a>, he examined SAP\u2019s approach. In this final installment, he focuses on the four key challenges – what is still holding companies back from putting agents into action?<\/strong><\/p>\n\n\n\n

There is clearly a lot of momentum behind AI agents at SAP \u2014 and strong customer demand as well. That raises an obvious question: Why is enterprise adoption still relatively slow?<\/strong><\/p>\n\n\n\n

Part of the answer may simply be that setting up and licensing SAP Joule for agent development is not exactly straightforward. But the real issue was perhaps best summarized at last year\u2019s SAP TechEd 2025:<\/p>\n\n\n\n

\n \n \n\n
\n\n
\n\n
How hard can AI agents be? Enterprise hard!<\/div>\n \n \n <\/div>\n\n<\/div>\n\n\n <\/div>\n\n\n

Once companies move beyond proof-of-concepts and hackathons, bringing agents into production quickly turns into a serious engineering challenge<\/strong>. The tooling, architecture patterns, and operational practices around agents are still evolving rapidly, and organizations are actively trying to figure out what reliable production setups should look like.<\/p>\n\n\n\n

Challenge 1: Human in the Loop<\/h2>\n\n\n\n

One of the most important requirements, not least due to regulations such as the EU AI Act, is maintaining some form of human-in-the-loop control<\/strong>.<\/p>\n\n\n\n

In enterprise scenarios, businesses typically want expert users to be able to intervene when needed, validate agent decisions, or participate in verify-and-approve workflows. AI agents should support human work, not silently operate outside of established governance processes.<\/p>\n\n\n\n

SAP is actually in a strong position here thanks to the integration between SAP Build Process Automation and Joule agents. Agents can trigger workflows, while workflows can also incorporate agents as automated steps within a broader human-centered process. This makes it relatively straightforward to combine AI-driven automation with traditional enterprise approval patterns.<\/p>\n\n\n\n

However, a key challenge remains: How does an agent know when to ask for help?<\/strong> Ideally, agents should recognize when they lack the necessary tools, data, or business context to complete a task and escalate appropriately. In practice, designing agents that reliably detect such situations is still an open challenge.<\/p>\n\n\n\n\n

\n
\n
\n
\n
\n

What\u2019s the Concept Behind Agents?<\/h3>\n

AI agents have taken the tech world by storm. Over the past year, systems that can plan tasks, use tools, and autonomously execute workflows have evolved from research demos into \u2026<\/p>\n

\n More about Agents<\/a>\n <\/div>\n <\/div>\n <\/div>\n
\n \"\" <\/div>\n <\/div>\n <\/div>\n <\/div>\n\n\n\n\n

Challenge 2: Evaluation and Optimization<\/h2>\n\n\n\n

Building an agent that works for a small, well-defined task is relatively simple. Building one that reliably performs across a wide variety of real-world scenarios<\/strong> is significantly harder. While prompt engineering and iterative tweaking can work for simple use cases, production-grade agents require a more structured build\u2013evaluate\u2013optimize cycle<\/strong>, similar to what we are used to from traditional machine learning projects.<\/p>\n\n\n\n

This, however, introduces new challenges. Effective evaluation requires:<\/p>\n\n\n\n