Operations | Monitoring | ITSM | DevOps | Cloud

A service desk manager launches a virtual agent in January. By March, chat conversations are climbing, ticket volume hasn't changed much, and the monthly report doesn't explain whether the investment is delivering value. AI rarely produces a single number that proves its return. The gains accumulate across thousands of support interactions, making measurement just as important as deployment.

Most teams can confirm their AI features are turned on. Measuring how often employees use them, which requests get resolved without agent intervention, and where AI is helping support teams work more efficiently is a different question. In InvGate Service Management, those capabilities live in AI Hub, a set of built-in AI features that includes the Virtual Service Agent, AI-assisted ticket resolution for agents, automated knowledge generation, and more.

Many “LLM observability with OpenTelemetry” tutorials stop at a single chat.completions span. That works for a demo, but it leaves gaps once an agent fans out into 30 tool calls, two vector-DB queries, three handoffs, and a 90-second tail latency you need to attribute. This post wires the OpenLIT SDK (50+ instrumentations, OTel GenAI semantic conventions, one line of code) into the full VictoriaMetrics observability stack and shows query examples that turn agent telemetry into decisions.

There’s a question we hear constantly from platform and engineering leaders right now, “which agent SDK should we standardize on for our Kubernetes clusters?” The honest answer is that the question is slightly wrong, and the rest of this post explains why. But it’s a fair question, so let’s compare the contenders first.

A year ago, AI coding tools worked alongside developers—suggesting the next line, completing a function, accelerating work that a human was already doing. Today, they’re writing entire modules and services independently, producing code that no human has reviewed line by line, built from components that no single person has fully mapped. And adoption is only accelerating: According to our recent AI Resilience Survey, 84% of organizations are now using AI to write, review, or suggest code.

Benchmarks keep showing that picking the right LLM is hard. The easy answer is "just use the most powerful one." That works, but it is pricey. A small, cheap, or local model can handle many simple requests just as well as a frontier model, for a fraction of the cost. That is what semantic routing is for. Use middleware that looks at an incoming request and decides which model should answer it.

This OpenAI API cost calculator (also an AI inference calculator for o3/o4-mini thinking tokens) estimates your monthly OpenAI API pricing bill from three inputs: model, request volume, and average tokens per request. Toggle between standard, batch, and cached pricing and get your number in seconds. It also shows what the same workload costs on Claude and Gemini. For the full per-model rate card, see CloudZero's OpenAI API pricing guide.