Jacob Ortega: Hi there, my name’s Jacob Ortega and I’m the Technical Account Executive here at Lydonia. If you’ve ever hired someone or worked in HR or the staffing agency, you’ve engaged in one way or another in the delicate balance of crafting or refining word descriptions for a company. With a variety of influences in both the industry, geographic location, required skill set, and job market in general, this task is not only cumbersome but, when done at scale, it’s a major headache for workforce solution providers.
With UiPath, we can automate and enhance this process at scale. We can perform dynamic analysis on submitted role profiles, enrich descriptions that in some cases are missing key details. Let’s see how an agent can handle a description like this that would be too ambiguous for us to upload to a major job site.
Something in the agent builder we can see the system prompt laid on in natural language, which gives instructions for the overall behavior of the agent. It shows the tools and contexts and specific instructions on how the agent should execute its analysis. The user prompt then shows more tactical information for how the agent should specifically handle the output of its response.
We then see the integrations for the specific tools that the agent will use: the web search functionality, the email communication functionality, as well as the RPA tool used to extract the initial job descriptions. We see a context grounding index, which provides vectorized snapshots of the agency’s internal database. It shows the agent’s successfully filled roles that it can use in its analysis, as well as an escalation tab in case a human needs to be involved.
We’re going to speed up the run time of the agent to be cognizant of your time. We’ll talk through the performance of the agent while it runs. So, we’ll see upon initiation the agent understands, based on its instructions, it needs to go and retrieve the job description in order to analyze it. Once the agent receives the description from the bot, it will query it against the context grounding index as well as any publicly available information through the web search functionality. The goal is to compile any of this relevant information for the enrichment, and finally we can see the output.
We see the original one-sentence description has been transformed into a professional, detailed role profile based on the existing information from the agency as well as external data the agent was intelligent enough to search for and determine the relevance of. Let’s take a closer look at that information.
You can see in the index there were several documents that contained information about similar roles and similar locations, giving the agent context for how to fill the gaps in the original role profile. We can see during the web search, the agent was able to query salary databases as well as public forums for better awareness before it rewrote the description.
We can make decisions with agents like this because of the UiPath platform capabilities. It allows us to test and ensure the decisions that we make with agents are not only confident but are tied and grounded to existing decisions, which essentially allows us to make these agents smarter over time. This agent sees more job descriptions and enriches more of them whether they’re partially complete, fully complete, or need an entire rewrite, this agent will become more intelligent over time as it makes decisions.
And because the platform is optimized for agents, we’re seeing real value for our customers, from increased accuracy to reduced manual efforts to overall process enhancement in use cases like this and more.
If you want to learn more about where we can help you in your automation journey, find us at lydonia.ai. Thank you.
