We delivered an AI endurance coach for a Charleston running store owner who coaches forty-plus marathoners and ultra-runners — one that manages individualized training plans across every athlete, adjusts workouts based on GPS-watch data and recovery signals, coordinates race-day logistics and taper protocols, and spots injury-risk patterns before they become six-week shutdowns

He opened the running store on King Street fourteen years ago and started coaching almost by accident. A customer training for the Charleston Marathon asked him for advice, then another, and within two years he had a formal coaching roster that now sits at forty-three athletes ranging from a fifty-eight-year-old woman running her first marathon to a twenty-six-year-old trying to qualify for the Olympic Trials. Every one of them has a different goal, a different injury history, a different weekly schedule, a different body, and a different relationship with pain. He was managing all of it in a spreadsheet that had grown to eleven tabs and a text message thread with each athlete that he checked between customers at the store, during his own morning runs, and at eleven o'clock at night when the West Coast athletes finished their evening workouts. His coaching was excellent. His system was failing. The spreadsheet could not tell him that an athlete's easy-pace runs had been creeping faster over three weeks — a pattern he knows from experience precedes either a breakthrough or a stress fracture, and the difference depends on context he could not see in a row of numbers. It could not tell him that two athletes targeting the same October race were both scheduled for their twenty-mile long runs on the same weekend and might benefit from running together. It could not tell him that an athlete who logged "felt tired" in three consecutive workout notes was also the one whose resting heart rate had risen eight beats per minute over the same period, which together suggest overtraining, not laziness. We built him a coaching system that thinks the way he thinks — in training blocks, not in calendar dates. Each athlete's plan is structured around their goal race: a sixteen-to-twenty-week macro cycle broken into base building, strength, peak, and taper phases. Within each phase, the system manages weekly volume, long run progression, workout intensity, and recovery days. But the plan is not static. Every morning, the system ingests overnight data from each athlete's GPS watch — Garmin, COROS, Apple Watch, or Polar, depending on what they wear — and evaluates the previous day's training against what was prescribed. If an athlete was supposed to run eight miles at an easy pace of 8:30 per mile and actually ran eight miles at 7:50 per mile, the system does not just note the discrepancy. It looks at the heart rate data to determine whether 7:50 was genuinely easy for that athlete on that day — because fitness changes and what was threshold pace in April might be easy pace in July — or whether they were pushing too hard, which the heart rate will reveal. If the pace was easy by heart rate, the system adjusts their training paces. If the pace was hard by heart rate, it flags the workout for his review and drafts a message to the athlete about staying disciplined on easy days. The injury-risk detection was what he cared about most, because a six-week injury shutdown does not just cost the athlete their race — it costs him a client. The system tracks seven indicators across a rolling fourteen-day window: training load change rate, easy-pace cardiac drift, sleep quality trends from wearable data, self-reported soreness scores, asymmetry in cadence or ground contact time, the ratio of high-intensity work to total volume, and the number of days since the last full rest day. No single indicator means anything. But when three or four trend in the wrong direction simultaneously, the system raises an alert with a specific recommendation — drop volume by twenty percent this week, replace Wednesday's track workout with a pool running session, schedule a visit with the sports medicine clinic on Meeting Street that he has a referral relationship with. In the first five months, the system flagged nineteen athletes for elevated injury risk. Fourteen of them adjusted their training based on the alert. Of those fourteen, none got injured. Of the five who did not adjust — because athletes are stubborn and believe they are the exception — three developed injuries that required time off. Race-day logistics were the part of coaching he liked least. When fifteen of his athletes are running the same marathon, they each need a different pre-race plan: different arrival times based on their corral assignments, different pacing strategies based on their goals, different fueling protocols based on what they have practiced in training, and different mental cues for the hard miles — because mile twenty in a marathon is a different experience for someone trying to break three hours than for someone trying to finish under five. The system generates individualized race-day packets for each athlete: a timeline for race morning, a mile-by-mile pacing chart with target splits adjusted for the specific course profile, a fueling schedule matched to the aid station locations, and a one-page race strategy that references the specific workouts they have done in training. When an athlete stands at the start line of the Kiawah Marathon holding a laminated card that says "miles 18-20 will feel like that tempo run you did on Sullivan's Island in August — you held pace then and you will hold it now," that is the system writing in his voice because it has learned how he coaches. The group dynamics were an unexpected benefit. When the system noticed that three athletes training for the same ultra-marathon were all within fifteen seconds per mile of each other on their long runs, it suggested they do their Saturday twenty-two-miler together on the Swamp Rabbit Trail. They did, they bonded, and they now train together every Saturday — which means they hold each other accountable and he spends three fewer hours per week managing their individual check-ins. "I got into coaching because I love watching people discover what they are capable of," he said. "I did not get into coaching to spend my evenings updating spreadsheets. Now I spend those evenings actually coaching — reading the data, thinking about the athlete, making decisions. The system handles the accounting. I handle the people."