When generative AI first entered the enterprise space, it mostly showed up in chat windows. Prompts in, answers out.
But the real opportunity, the one that turns AI into actual productivity, lies in agentic automation.
That’s what we set out to explore during the SnapLogic Agentic Bootcamp, led by Jocelyn Arcega. Four of our team members, Gjorge Argirov, Jovanche Angelevski, Antonio Ardzanliev, and Andrea Ristevska, joined the experience to learn what it really takes to move from prompt experimentation to production-grade automation.
Only two of them (Gjorge and Andrea) were available for a sit-down to share their impressions.
In this interview, they share what they built, what they learned, and where they see these new capabilities adding real value, for our team, and for our clients.
Why did you attend the SnapLogic Agentic Bootcamp, and what were your initial impressions?
Gjorge
Before the bootcamp, most of us were only familiar with language models like ChatGPT through basic browser usage. We hadn’t explored real-world agent implementations for the enterprise environment.
What surprised me most was seeing how an agent could do more than just respond to prompts, it could execute multi-step logic, access tools, send emails, even extract structured data. That practical integration through SnapLogic was eye-opening.
Andrea
Exactly. We all use ChatGPT daily, but this showed how to operationalize it. The AgentCreator was a functioning assistant that could carry out complex tasks. That distinction became clear during the bootcamp.
What kind of project did you build during the bootcamp?
Andrea
We built a real-time Jira Vulnerability Management Agent. The goal was to reduce manual workload and improve communication between SnapLogic’s security team and their customer support.
Instead of someone having to check Jira manually and email updates, our agent pulled ticket data, summarized status, and could automatically send those summaries via email.
Gjorge
Yeah, the use case was driven by a bottleneck in communication. Security staff had to respond to a high volume of queries about ticket status. We streamlined this through automation. The agent could fetch, summarize, and respond in natural language, saving time and reducing human error.
Was this just a concept or a functioning product?
Gjorge
The agent was live. We built it with real APIs, connected it to real Jira data, and used a frontend built with Streamlit to trigger agent flows. SnapLogic managed the logic, and the agent could choose which tools to use depending on the user’s prompt. So yes – it was real and functional.
Andrea
We presented it to SnapLogic staff at the end of the training, and the feedback was positive. It wasn’t just theory, we delivered a product prototype built for an actual operational need.
Do you see potential applications of this for current clients of IWConnect?
Gjorge
Definitely. One example we explored was with one of our clients we have at IWConnect in the financial sector. The idea was to build an agent that could assess creditworthiness. You upload a few documents – bank statements, income declarations, and the agent makes a decision based on predefined banking criteria.
Andrea
That project was still a proof of concept, but it showed how versatile these agents could be. From summarizing support tickets to evaluating loan eligibility, it’s just a matter of defining the workflow and integrating the right tools.
What’s next? How do you plan to use this internally or with clients?
Gjorge
Internally, we’re planning a full rollout of this tech in June, an implementation for our own workflows. We’re also considering using voice interfaces powered by ElevenLabs to make the experience even more intuitive.
I see a great opportunity in using your voice to just talk to an agent instead of typing prompts.
Final reflections—what was the biggest takeaway?
Andrea
The bootcamp demystified “AI agents” for us. It wasn’t about abstract theory, it was about building something functional in just 48 hours. We walked away with both confidence and practical knowledge ready to implement into our own use cases.
Gjorge
I agree with Andrea here. The main takeaway was about where do we see this implemented and applied. If you can define a process, SnapLogic + LLMs can automate it.
That’s the future we’re building toward.
Wrap up
SnapLogic’s Agent Creator didn’t just teach us how to build agents. It taught us how to think in tools, flows, and orchestration – how to delegate not just tasks, but entire workflows to systems that can reason, adapt, and execute.
We left the bootcamp with a working product, real technical confidence, and a clear path forward: Use AI not just as a brain, but as a doer.
If you’re a company looking to operationalize AI beyond the chat window, we’d love to share what we’ve learned, and help you move faster, with less friction.
