Smart Operator Service

Architecture Diagram

Inspiration

During a meeting in which a Customer Support Manager explained the challenges and difficulties that support teams have, when customers complain about something not working with their product, we immediately thought on how we could solve this issue. Bringing an AI companion to first line agents felt like a natural step towards the next generation of customer support teams.

What it does

Our Smart Operator Service is an AI tool that given a production Issue (can be Jira or any other system) is capable of analyzing the ticket, cross-check with the project documentation and with the solution components to find the root cause of the issue and produce a final report for the user in a couple of minutes, together with a detailed technical report with proofs of what cause the incident. The tool is specialized in the EV Charging domain, given we had the already existing solution infrastructure and documentation, but it can of course be applied to several domains and the AI core is not domain-specific.

How we built it

We built it leveraging mainly AWS AgentCore, for the reasoning and agentic part, and we used AWS services such as Stepfunction, Lambda, Dynamo, S3 and Cloudwatch to bring the solution altogether and finalize our project. For demo purposes, we also included a Jira board and a Confluence space to show the possible integrations.

Challenges we ran into

We had some challenges into connecting all the moving parts of the project, especially linking the whole flow end-to-end with Jira and Confluence components.

Accomplishments that we're proud of

We are immensively proud of the solution we have built, we are extremely happy with the results we obtained and we think this is a game-changer tool, in our every day working life. We are also proud of the quality of the technical and user-friendly report that our agent is capable of producing, knowing very well the complex dynamics of the sector.

What we learned

We have learned how to exploit AWS AgentCore for deploying a full solution based on AI agents. We learned how to integrate the solution with other AWS services, in order to deliver a full end-to-end product, ready to use. We learned how to tweak the agents prompts and different models in order for us to get to the best result possible.

What's next for Smart Operator Service

The sky is the limit, now that we know what our agents are capable of, we want to feed even more integrations in the system to improve the overall quality of the solution. We would like to explore the integration of the source code in the analysis, alternative storage solutions and ticketing platform to expand the coverage of our tools. Integrating the ticket similarity as part of the reasoning of the agents is also something additional we want to explore.