Jack Williams, iTero and GIANTX: The Commercial Boundary of AI Coaching in Esports

During the fifteen-minute break between game one and game two of a best-of-three, an analytics room has exactly one task. A machine-learning model pushes out a ban suggestion built from the opponent's behavioural patterns in the previous game. If that suggestion had appeared thirty seconds earlier, while the game was still running, it would fall squarely into prohibited territory. It appears during the break, when play has stopped — and during that break, the rules of most major leagues remain unclear. I sat with that detail for a while, because it explains why the Jack Williams interview about iTero deserves analysis rather than a quick read. The subject is not a new tool. The subject is a governance boundary that has not finished being drawn, and a market trying to price it. Jack Williams is behind iTero, an AI-driven coaching product. The partner on the other side is GIANTX, an organisation formed through the merger of Excel Esports and Giants Gaming, tied to the League of Legends ecosystem in EMEA. The interview carries two clearly titled sections: working exclusively with GIANTX and the likelihood of being copied, and AI-assisted cheating. For anyone who tracks league structure, those two headings are two halves of one problem. One half is commercial, concerned with selling exclusivity and protecting an edge. The other half is about integrity, concerned with which tools may be used in-game and how far. A third frame, fairness in the distribution of preparation resources, sits exactly between them and is barely mentioned. There is one notable chronological detail. In the biography section, the article recalls Natus Vincere lifting the Aegis of Champions at Gamescom fourteen years ago. The International 2026 took place at Gamescom in 2026. Simple subtraction places the piece around 2026. That is an arithmetic inference from the article's own wording, not a disclosed fact, so I mark it as medium confidence. More important: the article supplies no performance data at all. No sample size, no dataset scale, no evaluation methodology. For a product sold to professional organisations, that is the single largest gap. Numbers never lie — only the reader's heart turns them into lies. Here, the problem is that there are no numbers to read yet. Start with league structure. The LEC is a closed league. There is no relegation, and members are permanent members. In such a system, a structural advantage is not competed away over a season; it persists across seasons. If one organisation holds exclusive access to an analytics tool, that gap does not close itself. In an open circuit, other teams can climb and take the position. In a closed league, they can only buy an equivalent tool — if anyone is selling one. That is why I treat the central question as not whether AI cheats, but who gets to hold the tool. The article's two headings sit side by side as if they were separate topics. In practice they are one. An exclusivity agreement and an integrity rule both speak to the same thing: how much asymmetry in match preparation is permitted to exist. Next is the time window. Real-time in-game assistance is clearly prohibited across every major title, so nothing remains to debate there. The grey zone sits in the between-game window of a best-of-three or best-of-five, and in the pre-match phase. There, a model can read behavioural patterns from the previous game, cross-reference history, and produce adjustment suggestions within minutes. The game has stopped, but the match has not ended. The value of such a tool differs by title. I apply a decay coefficient to meta knowledge — measuring how quickly a tactical pattern loses value after each update. Dota 2 is operated by Valve on a sparse, systemic update cadence, with long stretches of stability between patches. A model trained on historical data retains validity for longer. League of Legends is operated by Riot on a two-week patch cadence. There, the half-life of any learned pattern is far shorter, and the value of AI shifts from solving the meta to detecting the meta delta faster than opponents. Those two titles reward two different kinds of advantage. One rewards depth of historical modelling. The other rewards speed. A product sold with the same promise to both is a signal that needs re-checking. This is a low-confidence inference, because the article names no title for the iTero product. The third point is publisher policy. Valve and Riot have different histories in how they treat third-party tooling and competitive data. If that holds today, an AI coaching vendor faces two markets of different size: one where the publisher is more permissive, one where it is stricter. I mark this medium-to-low confidence, because it needs verification against current policy rather than extraction from the article. The fourth point is copyability. The article has a section on it. Commercially, a tool built on models trained from public data has a lower barrier to entry than its surface suggests. Professional match data is largely published or collectable. The difficulty lies not in having data, but in turning data into a suggestion fast enough to use within fifteen minutes. That is an engineering problem, not a data-exclusivity problem. One methodological note. When a party announces an exclusive relationship without announcing performance, the only way to price it is to observe behaviour. Did the organisation restructure its coaching staff, add a data-analysis role, change its ban-and-pick approach in deciding games. Those signals are measurable and can be cross-checked between two independent sources. Based on my experience following matches, that is how I handle any information that lacks a baseline metric. And here I return to the starting point. A transfer is not the purchase of a person, but the purchase of a probability distribution. An exclusive tooling agreement is the same. It buys a distribution of advantage over time. If the meta's decay cycle is short, that distribution thins quickly. If the cycle is long, it thickens. Vendors have an incentive to choose titles and teams with long decay cycles, because there the advantage does not evaporate before the contract expires. Some matches end when the referee blows the whistle — and some only begin when the data speaks. The debate over AI coaching is one of those matches. It does not take place on the field, but in the governance room. The counterintuitive angle sits here. Both of the article's headings frame the issue around integrity and commerce. But the biggest risk an exclusive tooling agreement creates is not cheating; it is governance risk for the vendor and the league alike. If a tool influences competitive outcomes enough, the league operator will face pressure to choose one of two paths: mandate equal access for all members, or restrict the tool. Both paths erode the value of an exclusivity contract. This is a pattern that has already repeated with in-game coach communication rules: something once permitted, then progressively narrowed once its impact became clear. One more point: a team winning while using an AI tool does not prove the tool produced the win. Correlation is not causation. Separating the two requires a controlled trial, or at least a long sequence of matches with and without the tool. The article provides no data in that direction. Without it, every claim of effectiveness remains an untested hypothesis. And there is a paradox of value. If the tool is genuinely strong, it gets copied and the edge disappears. If it cannot be copied, it may well be because it is not that strong, or because the barrier sits in the contract rather than the technology. Both scenarios thin out the exclusivity story the headings imply. The signals to watch in the next cycle are three. First, whether league operators issue clear rules on access to analytics tools in the preparation phase. Second, whether the product is positioned as a multi-title tool or a single-title one. Third, the patch cadence of the title the target customer plays — because that is the variable that determines the decay coefficient of the very advantage being sold. I do not trust intuition — I trust the decay coefficient of intuition. For the AI coaching market, the practical question is not whether the technology works, but how fast the advantage it creates loses value before the rules are finished being written.