I’ve had similar experiences with the Ai ftp detection, I don’t do many group rides, i follow a plan.
I find it about 5-10 watts higher than expected, sure I could keep it where it puts me and “ raise my levels”. My main focus lately has been building TTE at ftp for a few months so a lower ftp gives me longer intervals. I’ve always been a more anaerobic athlete from years and years of bmx, so riding at threshold for a long time has been a struggle for me.
I thought about writing my experience to support about my issue but haven’t.
Have you reported this to TR support with specific numbers? E.g., what AI detection says your FTP is versus what you think it is? If not, you should. The only way it will get better is if people provide quantifiable feedback that the team can use to revisit / tweak it
My main reason for not doing this is that I know and recognize that the types of rides I’ve been doing would not result in very good or useful data for the system to come up with a reliable estimation, so I mentioned it here more for the sake of curiosity and discussion than anything. Perhaps if it’s still not accurate once I return to more structured training based on TR workouts I’ll raise it with support. They responded above with some suggestions anyway which is always appreciated!
Being selfish 
I think your use case is a great one for the team to look into. We all do unstructured outside rides, and these get factored into AI Detection. If AI Detection doesn’t handle these correctly, then it throws off the results for all of us.
So a case of “pure” unstructured rides is a good test case that the team could work with to improve AI Detections ability to “correctly” figure out the impact on FTP from unstructured rides
Yes, Post- Workout Surveys will have an impact on your AI FTP detection! This subjective feedback allows us to compare Training Stress against perceived exertion, which is useful information when determining your fitness level.
So yes, it will only be more accurate with more information! 
@SarahLaverty I probably missed it or mixed it up amongst the various discussions on AI FTP and AT, but will AI FTP auto adjust your FTP on it’s own if it seems off? Asking as I am returning from a prolonged, unplanned break. I would expect my FTP to bounce back to a point, however so far it has not prompted me to up it. I’ve been cruising through most of the workouts rating most as easy and seeing bumps in progression levels, however I did do close to two hours outdoors with an average and normalized power greater than my current estimated FTP a few weeks ago with no change. Do I need to manually adjust the FTP in these situations or just keep plugging until the next scheduled FTP test? Normally I wouldn’t expect a substantial change over the period of time I am talking about, but based on my situation leading into this block I would expect an initial quicker ramp in FTP as I return to training.
No. It will only auto adjust your progression levels, thus your workouts if they’re off.
Thanks. And to be clear, by auto adjust I meant suggest a new FTP for acceptance - not just make adjustments to it on it’s own.
As it is now, AI FTP detection works against progressing through PLs. I would like to work through progression levels, but I also love the FTP detection feature and would like to check and log those numbers as often as possible. As it is now, you can’t really do this because the detection (and the reversion back to your training FTP) messes with your progression levels.
It would be REALLY nice to be able to check and log this number without PLs adjusting or just have this number automatically logged in the background… or just be able to turn off the auto PL adjustment.
I realize that I am asking for additional luxuries here! But it would be nice to be able to regularly use the AI FTP detection and be able to progress further through the PLs. Thanks!
You can add your vote in the related feature request:
If you detect FTP but dont accept it then your PLs remain the same, no?
Correct. But at present, “logging” the data for any future review is a manual operation that must be documented by the user. I have used the Annotations in the TR calendar as a hack. The request above is asking for an integrated TR function for this tracking, even when a user rejects the FTP change (for reasons like pushing further into PL’s as one reason).
Yeah, it just seems to read, to me, that it affects PLs in any case.
I hear what you’re saying here. However, I think your suggestion could interfere with many athletes’ progress.
Progression Levels ensures steady progress and your FTP is an objective measure that reflects that progress. If you were to accept an increased FTP but keep the same Progression Levels, your workouts could increase dramatically.
The recalibration of your Progression Levels after an FTP increase is meant to keep a productive training progression based on your FTP increase.There is little benefit to testing your FTP more frequently than once a month when progress can be guided by evolving Progression Levels.
To get the most out of Adaptive Training and your Plan, I suggest testing or using AI FTP detection every 4- 6 weeks. That way, your progress will be sustained by working through the Progression Levels at a given FTP. After 4- 6 weeks, you can expect to see meaningful changes to your FTP. Subsequently, your Progression Levels will adjust accordingly to ensure progress at this new FTP.
Progress is always the goal! If you do want to log your FTP more often, you could do what @mcneese.chad has suggested; check your FTP without accepting it and log it using Annotations in your Calendar. Check out this article to learn how: Add and Edit Annotations.
Thanks for the reply @SarahLaverty ! Ramp tests and AI FTP detection tend to put me at a higher FTP that I cannot sustain for long, leaving me unable to train at adequate (IMO) time in zone. So while I like to check and log the number to see whether or not our AI overlords think I am improving…. or not, I then adjust the level back down to where I was and continue to progress through the PLs. You know what a manual adjustment downward down does to PLs. I understand why AT kind of “overreacts” or reduces PLs (rather than increasing them) when FTP is manually reduced, but in my case, PL reduction is simply due to me checking and logging AI FTP detection.
PLs should definitely adjust down when FTP is increased, but they also adjust down when it’s manually decreased. It would be nice, but it’s really not that big of a deal if I had the ability to log but not accept a detection. I can certainly manually select the next workouts and me and my progression levels back on track… just in time to check again.
@SarahLaverty Apologies if I have missed it in this big thread but when does the AI FTP D ramp test change to a workout? I adapted the plan pre Christmas (I forget if I added a time off to trigger it) and that generated a new Ramp/ AI FTPD. I chose the latter (AI FTP D). My originally scheduled Ramp/ AI FTP is tomorrow and of course when I select AI FTP D the current one is still valid. However, when I click OK the ramp remains; when does it get substituted for a work out? If it doesn’t no biggie I’ll just copy next Tuesday’s wo to this Tuesday. Thanks
I had this issue, and they told me to use the next Tuesday’s workout. Actually this time when I did AI FTP, but before AT had done it’s thing I briefly had the same workout this week and next week!
Hey @HLaB! I was able to prompt the replacement of your Ramp test with the appropriate Workout by heading to your Calendar, and clicking the annotation that marks the start of this new Block.
Once I clicked “Build”, the box below popped up. I hit “update” which refreshed the Plan, and ultimately replaced the Ramp test as expected. This is an issue that the team are working on, but this is an effective workaround in the meantime. Let me know if anyone has any questions about this! 
Quick question: Will AI FTP Detection work well with the beta Polarized plans?
Does for me!