The old model: plans that don't adapt
Traditional triathlon coaching has followed the same model for decades: a coach writes a plan, the athlete follows it, and adjustments happen in weekly check-ins. It works, but it has limitations.
Plans are static. Life isn't. You get sick, work gets busy, the weather changes, or you just have a bad week. By the time your coach adjusts the plan, you've already done (or skipped) several workouts that didn't fit.
The rise of adaptive training
Platforms like TrainingPeaks, TriDot, and Athletica have moved towards adaptive training — plans that shift based on what you actually do. This is a significant improvement. If you miss a workout, the plan recalculates. If you're fatigued, it might dial things back.
But most adaptive platforms still think in terms of plans — structured blocks of workouts that get shuffled around. They don't engage in conversation. You can't ask "why?" or "what about my knee pain?" or "should I race this weekend?"
The conversational AI coach
This is where AI coaching gets interesting. Instead of just reshuffling workout blocks, a conversational AI coach can:
- Understand context — not just your workout data, but your goals, injuries, schedule constraints, and preferences
- Answer questions — "Am I training too hard this week?", "Should I do a long ride or a brick session?", "How should I taper for my race?"
- Explain reasoning — instead of just prescribing, it can explain why a particular approach makes sense
- Remember — goals set three weeks ago, injuries mentioned in passing, race day targets
What makes good AI coaching?
Not all AI coaching is equal. Here's what separates useful AI coaching from generic chatbot responses:
Real data access — The AI needs to see your actual training data, not just hear you describe it. TRI-HARDER connects directly to Strava and intervals.icu to import your workout history, including metrics like heart rate, pace, power, and training stress.
Domain knowledge — A general-purpose AI doesn't know periodization from tapering. Good AI coaching is grounded in established training science — in TRI-HARDER's case, knowledge from Joe Friel's Triathlete's Training Bible is injected into every conversation.
Persistent memory — A coaching conversation that resets every session is frustrating. The AI should remember your race calendar, injury history, and stated preferences. TRI-HARDER stores coaching memory that persists across sessions.
Honest limitations — AI coaching should be transparent about what it is and isn't. It's not a replacement for a professional coach or medical advice. It's a data-informed training companion that can help you think through decisions.
The hybrid approach
The most promising direction isn't "AI replaces coaches" — it's "AI augments athletes." For self-coached triathletes (which is most of us), having an AI that understands your data and can have an intelligent conversation about your training is enormously valuable.
It fills the gap between "following a generic plan from a book" and "paying for 1-on-1 coaching." And because AI can process your entire training history in seconds, it can spot patterns and trends that might take a human coach several sessions to identify.
Where AI coaching is headed
We're still in the early days. Current AI coaches (including TRI-HARDER) are primarily reactive — you ask, they answer. Future iterations will likely be more proactive:
- Alerting you when your training load is trending into dangerous territory
- Suggesting plan adjustments based on upcoming races
- Identifying emerging patterns in your performance data
- Coordinating with wearable data for recovery and readiness
The triathlon coaching landscape is changing, and AI is a big part of that change. The question isn't whether AI coaching will become mainstream — it's how quickly it'll get good enough that every triathlete benefits from it.