# AI Workout Analysis

Canonical HTML: https://effortsignal.com/features/ai-workout-analysis/

Last content update: 2026-09-29

Understand what a completed workout involved, how the effort was distributed, and where it fits into your recent training. AI Insights turns the available workout data into a short, readable analysis.

<!-- The example is a persisted model response, not authored demo advice. Keep its exact facts
     in accessible text as well as real UI captures; translations are identified as translations. -->
<h2 id="example">A real ride, a saved analysis</h2>

This example comes from Warszawa Granfondo 2026. The analysis was saved on September 23, 2026. The screenshots and text below reproduce the same four observations.

<figure>

[![Saved AI Insights for Warszawa Granfondo 2026, with four observations about effort and training context](/marketing/product/catalog-ai-workout-analysis-desktop.webp)](/marketing/product/catalog-ai-workout-analysis-desktop.png)

<figcaption>The saved answer connects the distribution of power, the heart-rate response and the preceding training load. It also explains what the recording cannot establish.</figcaption>
<a class="marketing-image-enlarge" href="/marketing/product/catalog-ai-workout-analysis-desktop.png">Enlarge the AI analysis</a>
</figure>

### The four observations

- This was a long, demanding mixed-intensity ride rather than steady endurance work: 3 h 06 min moving, with weighted power of 175 W (84% of the saved 209 W FTP) and 216 PSS. Power ranged widely across zones—44% of recorded time was in active recovery, while 34% was in threshold or higher zones—so the average power of 143 W understates the harder efforts.
- The cardiovascular demand was sustained: average heart rate was 160.5 bpm (91% of the saved 177 bpm LTHR), with 162 minutes classified as vigorous by heart-rate reserve. Recorded exertion was 8/10, consistent with this being a substantial effort despite the lower average power.
- Aerobic decoupling was calculated as 8.5%, indicating that the power-to-heart-rate relationship changed over the ride. Because power was highly variable and only aggregate data are available, this does not establish a cause or show exactly when the change occurred.
- The ride added 419 TRIMP after three workouts totaling 584 TRIMP in the preceding seven days, including a 3,876 m swim the day before. The previous local day’s closing TRIMP load state had ATL 45.6 versus CTL 12.5; that describes preceding load, not readiness at the start of this ride.

<h2 id="context">What goes into the analysis</h2>

The workout comes first: duration, recorded effort, calculated metrics, intervals and laps, plus your notes and perceived effort where available. Power and heart rate describe different aspects of the session; PSS and TRIMP remain separate measures.

Recent training adds context from before the workout. The analysis can use the preceding 7 and 28 days, the previous day's training-load state and available recovery measurements. A historical workout is interpreted against its own dated evidence and saved thresholds.

When available, a linked planned workout and its saved execution results help explain how the session compared with the plan. Already-found similar workouts can provide comparisons with earlier sessions. Missing context stays missing.

<h2 id="open-analysis">Open it from your workout</h2>

In the web app, open a completed workout from Activities and select AI Insights beside its title. The analysis opens in a side panel, so the recording remains close at hand.

<figure>

[![The AI Insights action beside the completed workout title](/marketing/product/catalog-ai-workout-entry-desktop.webp)](/marketing/product/catalog-ai-workout-entry-desktop.png)

<figcaption>AI Insights is available directly from the workout header. On a phone, the analysis panel uses the full screen width.</figcaption>
<a class="marketing-image-enlarge" href="/marketing/product/catalog-ai-workout-entry-desktop.png">Enlarge the workout action</a>
</figure>

On first opening, a new analysis is generated if none has been saved. Reopening the panel reads the existing answer. Closing it while analysis is running does not cancel the request; return to the workout to read the result.

<h2 id="limits">Read the observations alongside the data</h2>

Missing sensor data or history limits the analysis. AI can help interpret the available measurements, but an observation is not a diagnosis or proof of a cause. The example's aerobic decoupling describes a change in the relationship between power and heart rate; it cannot establish why it happened.

Use the workout charts and interval details to examine an observation more closely. Saved answers reflect the data available when they were generated; reopening a panel does not automatically rewrite an earlier analysis.

[Explore workout charts and intervals](/features/activity-analysis/)

All Pro features are free through January 31, 2027. No credit card or automatic charges. After Early Access, keep using Free or choose Pro. [See Free and Pro plans](/pricing/).