How to get started with AI Analysis: Start making decisions from your load test results

#ai #performance #testing

Most engineering teams run load tests on a regular cadence. Far fewer act on them quickly, and it's usually not because the data is wrong. It's because there's too much of it. Response time percentiles, error rate curves, throughput breakdowns, request group comparisons: the signal is in there, buried under dashboards that take a while to read and longer to make sense of.

AI Analysis in Gatling Enterprise Edition is built to close that gap. It reads your results and tells you what matters, after every run, across your simulation history, and between any runs you compare. Here's how to get started.

What is AI Analysis?

AI Analysis is built into Gatling Enterprise Edition, and it works across three places in the product.

Run Summary gives you a read the moment a test finishes: what performed well, what regressed, what to look into before the next run. No manual pass, no scrolling through charts hoping to catch the anomaly.

Trend Analysis zooms out across a simulation's run history and tells you whether your system is getting better or worse over time. Is p95 creeping up over recent runs, or holding steady? Is that error rate a one-off spike, or the start of a real regression? The question your engineering manager asks next sprint, you can answer today.

Run Comparison is what you reach for after a deployment, a config update, or an infrastructure change. Pick two to five runs, click Analyze with AI, and get a report back in seconds: quantified findings on what moved, a verdict (Similar, Some Discrepancies, or Divergent), a confidence level, concrete recommendations, and a precise place to start in the comparison chart.

You can read more about all three on the AI Analysis page.

Step 1: Create your Gatling Enterprise account

AI Analysis is available in Gatling Enterprise Edition. If you don't have an account yet, you can start a free trial here, no credit card required.

Once you're in, an admin needs to enable AI at the organization level. You'll find the setting under your organization settings. Once it's on, AI Analysis is live across all your reports and dashboards with no further setup.

Step 2: Run your first test and read the AI Summary

After any completed run, open the run report. The AI-generated summary sits at the top of the page.

It covers three things: what performed as expected, what regressed, and what's worth investigating before the next run. It isn't a recap of every metric. It's a qualified read of what the data actually shows, laid out so you can act on it right away.

Already have runs in your account? You don't need to rerun anything. AI Analysis works on completed runs retroactively.

Step 3: Check your simulation's trend

Once you've got a few runs on the same simulation, open the simulation overview and look for the Trend Analysis panel.

This is where the AI reads your run history and tells you whether performance is heading the right way. Think of it as the answer to the question you should be asking before every release: are we in better shape than we were last week?

Trend Analysis is especially useful for:

Catching slow regressions a single run wouldn't flag
Confirming that a fix actually held across the runs that followed
Giving engineering managers a clear read on whether the system is improving sprint over sprint

Step 4: Compare runs after a change

In practice, this is where AI Analysis earns its keep. After a deployment, a config update, or anything else that might have moved performance, go to Compare Runs mode, select two to five runs, and click Analyze with AI.

Here's what the report gives you:

Findings: factual, quantified observations about what changed. Which request group regressed, where p99 moved, where throughput held steady.
Verdict: Similar, Some Discrepancies, or Divergent. One call on how the selected runs relate to each other.
Confidence: a reliability read, so you know how much to trust the conclusions. If a run was stopped early or the data's thin, the AI says so explicitly.
Recommendations: two or three concrete actions before the next run. Specific paths to investigate, logs to compare, config differences to review.
Explore in chart: a precise starting point in the comparison chart. Which metric, between which runs, where the divergence shows up clearest.

Step 5: Build it into your workflow

AI Analysis pays off most when it's a reflex rather than something you remember to check now and then. Three ways teams fold it in:

After every release: run a comparison between the pre- and post-deployment runs. The AI report joins your release checklist next to your smoke tests and SLO checks.
In your weekly performance review: open Trend Analysis on your critical simulations. Five minutes that tell you whether things are heading the right way.
When something feels off: instead of spending an hour hunting through charts, select the runs in question and let Run Comparison hand you a starting point. From "something feels off" to "here's what moved and why" in seconds.

What to expect

AI Analysis doesn't replace engineering judgment. It takes out the part that slows it down: the time you spend working out where to look.

The findings are factual, the recommendations are concrete, and the verdict is a place to start rather than the final word. What you do with all of it is still your call.

And that's the idea. AI Analysis isn't trying to make the decision for you. It's here so you've got what you need to make it quickly.