When I was at school, I once handed in a poem that I had written instead of my maths homework.
I did not get an ‘A’.
There’s no justice.
18 Likes
It’s no different to the ‘old’ AI detection in that respect, the prediction element just means you can see the effect of bunking off your training in advance rather than waiting until detection day.
9 Likes
That’s true of human coaches, too. They cannot know in advance that one of my kids woke up early and need a little love. Or that I am starting to get sick.
But you can plan accordingly. E. g. if you have other things that are higher on your list of priorities, then simply do less and leave yourself more room/time so that you can adapt more flexibly to unexpected events or can change your plans.
4 Likes
I think that’s exactly what it DOES do. If you skip/change/underperform, etc., it sees that and adjusts. That’s a good thing.
What you’re actually saying is that you wish it did not take into account that we aren’t all perfect and would pretend you followed the plan perfectly.
25 Likes
TR AI has completely changed my training for the better! I am (was?) one of those edge cases with a very strong anaerobic contribution and a very low FTP. It seems that Ramp tests and the previous FTP detection were vastly overestimating my FTP and training zones.
After struggling to complete 8min sweetspot intervals and only being prescribed 6 min O/Us for a few weeks on end, I grew concerned that my FTP was set too high. TR support had suggested I take a ramp test to confirm my numbers, but due to my nature, it gave me an even higher FTP! I simply could not complete these workouts and started reducing the intensity by 5%, to just barely scrape by each hard day.
Once I gained access to the beta a couple days before release, AIFTP dropped me from 218W 2.5W/kg (from ramp test) down to 195W 2.2W/kg (-9%). I was so relieved, even though it was a punch to the ego, I knew it was probably right.
Since then my workouts have been bang on! I have had tremendous success with the difficulty of my workouts matching the intended RPE. My VO2 work started out a bit easy, and with a “moderate” survey response, has quickly met me where I am at.
I did a batch of zwift racing for ZRL season 3 over January alongside 2 other hard interval days (definitely not optimal training) and saw a 1% increase to 197W. Which was small of course, however my workouts have been steadily progressing and I have been breaking into the 20min long SS workouts finally and with success!
Currently, my predicted FTP is 211W (+7%) in 16 days and 8 more workouts! I am actually seeing my week over week progression in my wattage, and I have been super motivated to follow the plan more closely now that my workouts aren’t destroying me every time!
I think I am finally targeting the correct zones and intended stimulus and that is leading to actual improvement. I previously felt like I was bashing my head against the wall, working so hard for little to no improvement. It seems obvious in hindsight now, but I was clearly using an overestimated FTP before, I just wanted to trust the system. Now I can, and am excited to see where I can get to before my first gravel race in July!
This has been long winded, but I need to also mention that I have been seeing these improvements alongside some weight-loss going from 86kg to 81kg in the past 5 weeks! My goal has been to achieve 3W/kg for a long time, and it truly feels like I will get there this year.
My calendar if you are interested:
17 Likes
you are blaming TrainerRoad for biology?
3 Likes
For me the AI predictions have been. Pretty decent but I’m not following a Trainerroad plan. I do mostly 1,5-2h Z2 work and one VO2Max session / week and that seems to work pretty well for me personally. I’m actually using the figures Trainerroad AI is giving me and that also seems to be accurate.
1 Like
Add an activity called pump jump berm bump. Add your working sets. Or just use strength training. Add upper body sets and reps. Your downhill workouts will be accounted for.
I just received my first AI FTP Detection today after the new prediction became available. It jumped from 256 to 275—which is within ±2 points of what the “Early Detection” was suggesting throughout the block. It’s incredibly motivating to see the progress reflected in the numbers!
I wanted to share a few observations for the athletes out there who are actually time-crunched. (I always have to chuckle when podcasts refer to 6–10 hours a week as “time-crunched!”)
My “Minimum Effective Dose” Approach
I currently don’t have the bandwidth for a high-volume plan, so I’ve had to be surgical with my training:
-
Frequency: 3 (sometimes 4) rides per week.
-
The Mid-Week: Two 60-minute workouts (Tuesday/Thursday).
-
The Weekend: One 60–90 minute structured workout OR a semi-structured outdoor hill ride (120 minutes).
-
The “Secret Sauce”: I prioritize a full rest day between every single effort. This spacing, paired with focused nutrition, allowed my Workout Levels to progress consistently without burnout.
Ditching the Plan for AI Flexibility
Instead of following a rigid Training Plan, I’ve been using the AI Workout feature to maintain flexibility:
-
I played around with different zones to see which gave the “most bang for the buck” regarding predicted FTP.
-
It was eye-opening to see the immediate impact (positive and negative) of adding commutes or shifting days around.
-
The AI almost always nailed the “Hard/Borderline Very Hard” intensity, which is crucial when you only have 3–4 hours a week to make an impact.
The Takeaway
It’s the same old song, but it works: Consistency + Structure + Recovery. For those of us who can’t hit those 10-hour weeks, the AI features make it so much easier to maximize the limited time we do have. A big THANK YOU to the TrainerRoad team for building tools that work for athletes with “real world” schedules!
12 Likes
Or those or us who don’t WANT to ride 10 hours a week. 
I like TR for not making me do FTP tests 
and providing me the flexibility to throw in a random Zwift race or keep my hard group rides. It just removes the friction from getting the workouts done. Even if I’m not on a plan and just want to pick from TrainNow workouts.
5 Likes
Totally agree with everything in your post. Just a comment on how discouraging it is that today’s “real world” means people do not have time to exercise and self care. 
2 Likes
I agree with your overall sentiment you can be very fit doing far less than 10 hours of exercise a week. But overall it’d be nice for most people to have a better work life balance to allow them to actually have more of a choice to do 10 hours of cycling a week.
3 Likes
Revisiting this…
My schedule, as noted before, is highly variable and I often have to move stuff around due to work and life challenges. I’ve been VERY happy with the workouts since I joined the beta; workouts have been awesome and I feel like I’m getting stronger as I work through the plan.
I’m bad about adding in longer unstructured rides since, well, that’s what I’m training to do, after all, and riding is more fun than training. Even with this, I’ve seen slow steady progress, with AiFTP gains slightly greater than what I’d been seeing in the Fall with adaptive training and, again, many unstructured rides. I’ll be 61 in a few months and, frankly, I simply don’t expect to see large gains over short periods ever again. Single digit FTP gains over a four week period are plenty!
I did a ramp test a couple of weeks back just before my AiFTP detection just out of curiosity (on zwift, so it’s “just a ride” to TR-that was deliberate) and got the highest number I’ve ever tested to, which was a nice confirmation of how I’m feeling on the bike in the real world. The ramp FTP and my AiFTP do not align, with the Ramp test being significantly higher, but for me these are just numbers. I completely fail to understand how folks think that just because a number went up or down overnight, in a computer program, that they somehow became more or less fit than they were the day before.
It’s a number, folks.
Anyway….last week was a total disaster at work and I got several 24 hr call shifts with 3 hours or less sleep, and just didn’t feel like I could execute my planned workouts. So, on my Monday day off, I went for a ride that I’m quite familiar with and decided that I’d work on actually applying some of my newfound fitness on the ride.
Result: multiple PRs, with several updating from 2015/2016-a decade ago, and a time when I was VERY fit.
That was so much fun that after a terrible week of little sleep and no structured work, I went for a ride on Saturday (yesterday) that I’ve been riding for years, with the same goal: See how the strength I’ve been working on translates to the real world.
Result: Multiple PRs, again with at least one updating from 2016.
That was pretty good and the weather is awesome, so I went out early this AM for a predawn ride, and yep-several more PRs-and today was supposed to be a little easier than yesterday.
So. Maybe, if the numbers are bugging you that much, go for a ride. Just sayin.
SA
22 Likes
This is a fantastic post! 
1 Like
I know what you mean, but it made me smile you quoting Zwift as being the “Real World” 
2 Likes
I’ve kept quiet about the areas where I’ve had challenges and as expected, the system has dialled in.
It took 6 weeks but am now getting Threshold & O/U that are the best I’ve been prescribed for years - they were likely too hard on the previous version too.
I was overall happy 6 weeks ago but even happier now.
3 Likes
Ive been happy with the AI thus far and slowly (confidently) making progress which imo, it should be. There should be no massive jumps so I am happy when I see a 2% or 3% gain.
However, I have an AI FTP cycle due next week, 3% gain predicted which is great. But some of the workouts which follow look nuts compared to my 6 week PB’s.
Namely Truchas+2 & that 45min-60min section:
Will try as always, but it effectively looks like doing back to back 23min TTs at almost max effort with a 3min break inbetween. Fingers crossed
2 Likes
That’s not at all what I said.
I said that the Zwift ramp test confirmed my impression of how I’m feeling out on the road and the trails-the real world. Sheesh.
SA
It is your 6 week max. Highlight over it and it might be from another workout. So the set of PRs might just represent a workout with a little less rest or 3% more power.
Nope. You’re not the only one. In point of fact, AI is so good at prescribing workouts of appropriate difficulty that when my workout seems a little easier or harder than I thought it might be, I assume something is different with me, and not TR.
1 Like