The purpose of Agentic AI is to assist engineers excel in the job. The engineers are either into software development and testing or supporting the product. We have software coding assistant agents for developers widely available in the market. Do we have similar AI agents for support engineers? I would say this space is still evolving.

Quick look at AI agents in Software Development:

AI agents are trained in software languages like Java, Python, NodeJs and given a user query in English, agents can generate the code. Developers can quickly make use of the code snippet and integrate the generated code in the end to end system and accomplish the task.

AI agents in IT Operations:

Observability helps in collecting the health of a software system. There are AI solutions available to predict the abnormal behavior of the systems and notify the admins on possible failure scenarios. Then admins look into the generated alerts and take the necessary action to handle the possible failure.

For example, disk storage is reaching more than 80% of the threshold for an hour, and on alerting the admin about disk usage, admin can take action like adding more disk capacity to avoid the out of storage capacity failure.

So, prediction of possible failures help IT administrators to an extent. The other possible scope of scenarios where AI agents can evolve in IT operations that assists both the IT administrators and the support engineers of the software system is to resolve the alerts on its own.

1. Root Cause Analysis(RCA): AI agents can also monitor the system metrics, traces and logs for every generated alert from Observability models and present the summary of analysis on what could have caused the alert. It assists the Administrator in identifying if the alert is true positive and in taking an action on the generated alert.
2. Automated Decision Capability(ADC): AI agents can monitor the actions taken by administrators on the generated alerts from Observability models. And develop the pattern on making a decision using the RCA and past decisions from the Administrator.
3. Execution: AI agents can be equipped with multiple tools that can perform the set of actions on a software system. Then AI agents can invoke the corresponding tools to perform desired action.
4. Reaching the Product support team: There are scenarios where Administrators need to open the defect and engage the support teams from the software vendors, AI agents can assist with these repetitive steps of opening the tickets, collecting the metrics, traces and logs and summarizing the scenario.

We have seen how AI agents can help the IT administrators. The other side is how AI agents help the Support Engineers?

Let’s look at AI agents role in assisting Support engineers on receiving the defect at customer site.

1. Analyzing the defect: Goes through customer provided summary and collects initial diagnostics. The list of items needed for a support engineer can be packaged into an automated tool that can collect required diagnostics, then an AI agent can trigger the tool and collect all the required diagnostics for a defect.
2. Search in knowledge base: The past resolved defects knowledge base can be used as a source of information and train a RAG model that assists all the support engineers to identify if the defect is already known.
3. Reaching out to SME(Subject Matter Experts): Support Engineers reach out to Development engineers on defects that cannot be resolved by them. An SME can guide the support engineer where to look further in resolving the defects. AI models can be trained to parse all the metrics, traces and logs collected and can provide actionable insights and eventually mature the models based on monitoring the SME recommendations. This is like how coding assistants help in generating the code.
4. Work with customer on further diagnostics: All possible diagnostics like enabling debug logging, collecting memory dumps and running any scripts to collect additional data can be packaged as multiple tools and let AI agents decide on what additional tool can be used in collecting further details based on the problem.
5. Coming up with Possible fix: AI models can be trained on understanding additional diagnostics and suggest actionable insights that help the Support engineers in taking necessary steps in resolving the defects.

This summary helps in developing the AI agents in IT opertions that assist both the IT Administrators and Support Engineers of a software product and makes the life of support cycle easy. Happy to hear your thoughts in the comments.