Should you trust your training to AI?

It’s worth considering a few things before taking the plunge.

AI will have an incomplete data set

Unless you record ALL your metrics that have a bearing on physiology like walking, washing the car and gardening the data AI has to work with it is incomplete so, in the absence of complete knowledge, it cannot accurately plan intensity, frequency and duration of training.

AI only measures the measurables

HRV and sleep monitors are a poor proxy for recovery. Heart rate is useful but variable. FTP settings are critical for training zones but will fluctuate and will differ between turbo and road training sessions and between bikes, particularly aero, race, gravel, MTB and, if you use one, a winter bike. AI doesn’t know the difference. Additionally, none of these measurements tell AI how you feel. If you could find a way to tell it how you feel about your short term commitment, long term goals and what motivates you, it would not know what to do with the information. It has no way of interpreting it into a psychologically a sustainable training pan.

AI uses Training Stress Scores (TSS) to plan training

A 10 mile time trial is about 35 TSS. A 4 hour Zone 2 ride is about 210 TSS. If you did 6 10 mile time trials throughout a week, your fatigue level would be completely different from a 4 hour Z2 ride but the Training Stress Score would be the same. AI cannot detect the difference, but your legs would.     

AI cannot know you

Try as it might, AI cannot reflect individual psychology. It does not know your personality type, behavioural characteristics and what you do and don’t like. It does not know when your mind needs a break or what encouragement you might need to re-engage in training after being floored by something out of the blue.

AI does not ‘do’ accountability

You cannot be accountable to a machine and it cannot be accountable to you. Machines don’t care and they don’t understand. It’s not possible to develop a healthy relationship with computer code.

More often than not, any training plan is better than none so AI might be worth a try - it’s also going to be cheaper than hiring a coach. Or at least it should be. However, a plan tailored to your individual psychology and physiology, reflecting what you want to achieve as a rider is likely to be more effective, sustainable and satisfying. If you do decide to trust your training to AI, I'd be fascinated to hear about your experience. 

---

​
​Rich Smith is a psychology graduate and a British Cycling qualified Level 3 Road and TT coach supporting riders nationally and internationally. He established RideFast Coaching in 2015 to provide effective training to cyclists after becoming convinced the machines are trying to kill us all.