Choosing the right AI capability

Choosing the right AI capability for your custom application is important. Learn about the differences between skills, AI agents, and agentic workflows, use cases for each, and when to choose one capability over another.

When adding AI capabilities to your custom applications, the capability that you choose depends on several factors, including the complexity of the workflow and the degree of human oversight required for the tasks.

When to add skills

Skills are suited for specific, well-defined tasks, such as summarizing a record or classifying a request. Because skills can be reused in different contexts and applications, skills are ideal for situations where the input and output are predictable and don’t require complex decision-making. Skills also provide a high degree of human oversight and control, which makes them a practical way to introduce AI capabilities within your organization’s workflow.

When to add AI agents

AI agents are suited for more complex, specific tasks that involve some degree of reasoning, planning, and responding. For example, an AI agent for a retail company app could be trained to identify recurring issues with product merchandise and escalate issues to the appropriate teams responsible for resolution. Because AI agents can contain multiple tools and skills, they are useful in situations where you want an entirely automated solution for a task or to offload manual work. AI agents involve less human oversight than skills, while still providing transparency and human intervention when necessary.

When to add agentic workflows

Agentic workflows are suited for complex, knowledge-intensive tasks where the solution requires human oversight at several steps in the process. Because agentic workflows can complete processes from end-to-end, agentic workflows are useful for processes that are low risk, such as handling routine approvals or requests.