One of the greatest questions of our time is how many jobs AI will make redundant in the coming years. One job very much in the firing line is triathlon coaches with the appearance of AI training apps which claim to be able to replace a coach, if not work better than one.
Initially I was extremely sceptical about this, but then I softened my approach. There is nothing we can do to stop the advancement of AI, so should we embrace it to automate the more mundane aspects of our coaching business? The writing of the plan itself isn’t the most stimulating aspect of the job, as the human interaction element is far more impactful than dragging a workout into the calendar, tweaking it for the athlete and moving onto the next one. If AI could take care of that, would that free me up to spend more time on what matters?
So I decided to open a coaching and athlete account with a view to embracing this new age of technology. I wanted to like it, heck, I was excited to try it. So before you read this rest of this article I want you to keep that in mind. I will of course be keeping the name of the software I used anonymous.
The first red flag was when I was setting up my account. It asked me if I wanted to take a rest day, and on what day I would like that to be. I said that yes I would like a rest day, and I would like it to be Monday.
If I was working with a coach I would expect them to simply nod, and make a note of that. They asked me a question and I gave an answer. Then an error message came up. I was recommended against having a rest day, as for my athletic development it was important for me to train for seven days a week.
Sorry, what? I had already told the software I’d been training for 12 years, but it was implying it knew better than me what was right for my body? It pushed back hard against my rest day, but relented eventually.
I also had to “unlock” strength training. That’s right, something as fundamental to triathlon training as strength work was a feature that didn’t come as standard. This as much as anything else was a major cause for concern.
I was also asked what day I’d like my long bike sessions to be on, connected my Garmin account and my first week was generated.
It also asked me to input a race, for which I chose a middle distance race in June, when I started training in early January. This is an important note for later.
My first week was then written for me. There were seven sessions, only one of which was an easy session, which was after an hour long threshold session on the Saturday, my “long” bike session.
This seemed off to me for a couple of reasons. Firstly, an hour isn’t an especially long bike session. Secondly, there was a LOT of threshold work in there, around 35 minutes worth. Remember, I am training for a 70.3 in June, and this is January. Threshold work will improve my FTP and muscular endurance, but keep in mind I had connected my Garmin account, and have done no threshold work for several months. The AI looked (or didn’t) at my training history and decided this was the right session for me. I decided to trust the process and attempted the workout, but blew up after the first interval, ended the workout and rode zone 2 for the rest of the workout.
Then there was the issue of a brick run. I told the software I lived in the UK, it knew it was January, and I had been in the sport for over 12 years but still set me a brick run. Why? It’s not like my legs don’t know how to run off the bike, why do we need to do brick runs six months out from an event? In the middle of f****** January? What am I supposed to do? Peel myself off of the turbo trainer, have a shower, get changed into winter run kit and then go out for my run in -2 celsius? Instead I did the 20 minute run in the evening to stick to the overall volume, which was quite pleasant in the end, even if it wasn’t what the software wanted me to do.
The rest of the week was broadly ok, but it was tough. Was I just a bit soft? Had I just been away from structured training too long and forgotten how hard it should feel? I decided to suck it up and keep going.
When week two came around I was still pretty happy to train, but the second week was… strangely similar to the first. Now, I appreciate that the best training isn’t always the most varied, what concerned me more was the repeated threshold work nearly every day. The only easy sessions was brick runs, the sessions weren’t overly motivating, and I did start to fall off the bandwagon. I woke up some mornings just not wanting to train, too tired and drained.
I did make it through the (marginally) easier threshold session on the Saturday, but I was completely cooked after, and couldn’t face training on Sunday. It was raining and cold so I also sacked off the brick run. Are they working on the assumption everyone has a turbo trainer and treadmill setup?
The wheels came off during this week, I only hit two sessions in the week, and failed my threshold session on the weekend. The AI made absolutely no attempt to make things easer for me, it was repeated high intensity sessions. Many of the sessions had threshold intervals at the start then zone two for the last 35-40 minutes, but the threshold work will fill the legs with lactic acid, reducing the contribution from the aerobic system as the body recycles lactate, which blunts the response.
It also asked me to do some very strange sessions, such as one where I needed to ride my bike, climb off, do some squats, some sprints, more squats e.t.c. which I honestly just ignored. This was likely a ham fisted attempt at triggering the post activation potentiation effect, but it felt like a novelty to me.
What was frustrating for me is that there was no way to give feedback or any other user input after sessions. If after every session I was able to enter an RPE value, that could help inform the AI of when I was struggling and needed less volume/intensity. Instead, there was no way to tell the AI whether I missed the session because I was exhausted, because I was visiting my girlfriend that Sunday or I just couldn’t be arsed.
I was hoping that an AI programme would adapt around me to create something bespoke for me, but the opposite was true. It had a very stubborn idea of how I should train, and was absolutely adamant I was going to train that way.
In the sales pitch I was told that it was a very objective way to train which didn’t fall foul of “trendy training philosophies”, which reeks of arrogance to me. The idea was that the AI knows best and that I need to learn to respect that. The trendy philosophies it refers to are actually well supported by science and have been used by elite athletes for decades, where the philosophy pushed by the AI seemed to be to fit in as much threshold as possible, as that’s what makes people faster. Right?
It reminded me of the illusion of choice you get as a child, when your parents ask you what you want, when they’ve already made up their mind. Rather than the AI looking at my training history, my goals, equipment available to me, weather and all the other areas a human coach would consider, I was instead put on a set course, and course corrected if I strayed from that. It felt like the AI was constantly trying to lure me back to doing what it wanted me to, rather than adapting my training based on the data I uploaded. I have a sneaking suspicion that the entire AI is based solely upon the adherence score you achieve at the end of the workout.
The result? Burnout. I just couldn’t face training, and in the last few weeks have only done a couple of sessions a week. The AI programme chewed me up and spat me out. I’m a professional coach, I live and breathe triathlon, for it to take the wind out of my sails so dramatically really takes some doing. I was so driven and optimistic at the start of the year, I now feel broken and demoralised.
I have chatted to others about the programme in question, many who are clients that came to me after trying the software themselves. They all said the same thing, that the training was too intense, and it left them feeling burnt out. The real kicker? A full subscription to the AI software in question costs TWICE the cost of my entry level package, and marginally more than my monthly coaching package. That’s a lot of money to spend on repetitive workouts and a programme which tries to railroad you towards a set destination.
If you’re an athlete reading this, I’m not saying you shouldn’t try it, it clearly works for some people, but the value it offers relative to the cost doesn’t add up for me. If you’re looking for structure in your training and something that will see you as an individual, hire a real coach. If you’re impressed with what AI can churn out, I will blow your mind with what I can do for you for less money,
If you’re a coach reading this, we don’t have anything to worry about in the foreseeable future. There is so much we can offer an athlete that AI is nowhere near being able to replicate. I believe that before the end of my career as a coach I will be working alongside AI in some shape or form, but it won’t be any time soon if my experience is anything to go by.