Some areas in Oracle are configured once and rarely touched again. Journeys are not one of those areas. They may require regular updates and additions as the organization grows, and when those needs come up, HR teams can often feel lost on where to start. That is what led me to build an AI agent that guides them through the process simply by using a chat window. My goal for this project was to learn the ins and outs of building an AI agent, and it proved to be a very fun challenge.
What I Built
The agent creates Journeys and Journey Tasks through a conversation. You start by defining the Journey itself, then walk through the tasks that make it up, including task type, performer, initiator, expiration, and due date. If you already have a Journey Template ID, you can also use the agent to modify an existing journey rather than starting from scratch. The goal was to make something approachable for someone who does not live in Oracle every day.
What Surprised Me
The current out-of-the-box options for AI agent creation are more limited than I expected. Most pre-built tools are designed for fetching data, not making changes. If you want the agent to actually do something in your system, you are building custom tools and explicitly defining every field it can access. The bigger the scope, the more complex that gets. Each element added is an opportunity for the agent to break, and I learned that the hard way.
The context awareness was a pleasant surprise. Asking for five onboarding tasks returned results that felt like they came from someone who understands HR. It picked up on implied differences, like new hire versus manager onboarding, without me needing to spell that out. That was genuinely useful.
Consistency was a different story. The agent’s delivery shifted from session to session. Functionally, it was performing the same tasks, but the way it was asking me questions was always different. That required more tuning than I thought.
The last thing that caught me off guard was how useful AI was in the building process itself, not just the end product. Writing the agent’s instructions, handling API calls, and making sense of error logs. Things that would have slowed me down significantly before became much more manageable with AI involved.
What This Means for HR Teams
This does not replace HR judgment. It removes the part that slows everything else down. A lot of time spent building a journey goes toward figuring out where to start. This skips that part. You describe what you need, get back something usable, and edit from there.
Closing Thought
Building this opened my eyes to what an AI agent is capable of. It made me wonder how far this could realistically go. Is there a future where entire Oracle modules are configured through a chat window? Maybe. The technology is closer to that than I would have guessed before starting this project. The harder question is where the sweet spot is between AI assisted configuration and professional implementation. Some things benefit from the speed and accessibility AI brings. Others need the judgment and accountability that comes with a real implementation. Finding that line is going to be one of the more interesting conversations in this space over the next few years.
