Jack Williams, iTero and GIANTX: AI Coaching Is Touching a Grey Zone Esports Refuses to Define

For eight minutes between game three and game four of a BO5, there is a room no league has ever fully defined. The coach sits there with three data layers open: a position heat map, an estimated cooldown table, and a machine-generated panel claiming the opponent wins 63% of fights when vision is placed on the lower half of the map. He is allowed to read it. He is allowed to write it down. He is even allowed to tell his players the opponent prefers late fights. What sits outside the permitted zone is something else entirely: exclusive ownership of the tool that generated that panel. Jack Williams' interview on iTero, GIANTX and the future of AI coaching in esports puts a finger precisely on that spot. And like every grey zone in this industry, it gets discussed as a technology story when it is really a governance story. Two section headings disclosed from the original piece carry nearly the whole focus: one on working exclusively with GIANTX and the likelihood of being copied, one on AI-assisted cheating. Between those two sits the gap I believe will become the loudest topic of the next two years: whether an exclusive commercial arrangement creates an unlevel playing field inside a closed league. Let me be direct about the limits of the source. The interview does not state Jack Williams' formal role, does not publish iTero's evaluation methodology, provides no sample size, no verification metric and no patch reference. Every performance claim in it is unverifiable from the available material. That does not strip the interview of value. It changes the kind of value: what is worth analysing here is the structure of a forming market, not numbers nobody has checked. To see why this grey zone matters, split AI coaching into four time windows. First, pre-match preparation: opponent analysis, draft planning, lane scripts. Second, the draft itself, where coaches remain permitted to interact in most competitions. Third, the break between games. Fourth, live in-game assistance. The fourth layer has nothing to argue about. It is clearly banned in every major competition, and has been for years. If the interview devotes a section to cheating, its subject almost certainly sits in the third layer — the between-games window, where the rules are blurriest, where a model can ingest data from three games already played and return a tactical adjustment before game four begins. That window lasts a few minutes. A few minutes is enough to change a BO5. Based on my experience watching matches, across both Asian regional League of Legends competitions and Dota 2's The International, one detail catches my eye that broadcast rarely shows: the coach-room camera. In 2026, most screens there were paper drafts and a replay player. At recent events, the number of data windows open simultaneously on a coach's machine has risen noticeably, and the tab-switching speed is well beyond what a human can read. That is the signature of a shift: from display tools to recommendation tools. The core point is that a recommendation tool does not sell knowledge. It sells speed. And speed only has value once the opponents' knowledge baseline is saturated. To see how title-dependent this is, compare two patch cadences. Valve runs Dota 2 on large, infrequent, disruptive system patches with long stable stretches between them. Riot runs League of Legends on a two-week cadence, each touching a few numbers but accumulating into continuous meta drift. The consequence is concrete. In the slower-patching title, a model trained on historical data retains validity over a longer window, and its edge is depth of modelling. In the two-week title, the half-life of any learned pattern is short, and the tool's value shifts from solving the meta to detecting the meta delta faster than opponents. The edge changes nature: from knowing more to knowing sooner. A product marketed identically across both title types is a red flag. I say that as a structural inference, not as an accusation against iTero, since the available material does not say whether iTero pursues that strategy. What the material does give us is GIANTX. That is the only organisationally verifiable anchor, and it opens the second analytical layer. Industry reporting indicates GIANTX was formed through the merger of Excel Esports and Giants Gaming, operating in the EMEA ecosystem with an LEC slot. I flag this as requiring independent verification, since organisation naming in the original piece can be misleading. But if accurate, the governing framework for a GIANTX–iTero arrangement is Riot Games' third-party software and competitive integrity rules. This is where it gets interesting. In a closed franchised league, structural advantages are not competed away. They accumulate. In an open circuit with promotion and relegation, a team holding a better tool gets imitated or removed by performance pressure. In a closed league, every participant is a permanent member, nobody leaves for losing, so analytical resource asymmetry persists across seasons. It does not flatten itself. An exclusive analytics agreement sits, by nature, in the same regulatory category as any other competitive-preparation advantage. A league permitting it is choosing to permit preparation inequality. That is a governance decision, not a technical accident. History shows organisers always respond in the same sequence. First they allow, while the tool is new. Then they experiment with expanded coach-player interaction during play. Then they narrow it once they realise the advantage is not evenly distributed. Finally they must choose: mandate equal access, or restrict the tool. This is why I consider the exclusivity-and-copying section of the interview the more important one, even though the cheating section generates more noise. On cheating, separate two questions. The first is whether a specific act is prohibited. The second is whether organisers can prove that act before concluding. The second is far harder, because language-model and statistical-model outputs carry no clear fingerprint. When a coach tells his players the opponent wins fights on the lower half, there is no way to distinguish a machine suggestion from a human read. A cheating accusation built on communication content hits that wall immediately. This is an asymmetry worth naming. The cost of a false accusation is far lower than the cost of proving innocence. Every wave of unverified accusation burns credibility for both the accused and the investigators, and next time a true accusation is harder to believe. A governance system is only healthy when it can handle both error types: missing a violator, and convicting an innocent. So far, this industry only has experience with the second. Now the part many in the industry dislike hearing. The copying risk the interview raises is the wrong question. The moat of an analytics product is not the model. The model can be rebuilt in months from the same public data. The moat is the data pipeline, exclusive data access, and how deeply the tool is embedded in a team's daily workflow. Copying an algorithm is easy. Copying an operating relationship is not. If an AI coaching vendor worries mainly about its model being copied, that signals it has not built a real advantage. The bigger, less discussed risk is the effect on the coach development pipeline. Match analysis is a skill forged through thousands of hours of replay review. When that layer is handed to a tool, the next generation of analysts never gets to build foundational skill. They receive the answer before learning to ask the question. I have held this view for a long time, independently of AI. Investment in grassroots coaching development in esports was already thin. Many academy programmes were opened with commercial motives first, while junior analyst roles barely have a career path. AI coaching enters exactly when that structure has not yet formed, and it will make building it harder, not easier. Legends do not die of mistakes. Legends die of data that knows how to count. That holds for teams, and it holds for an industry illusion: that esports is a pure meritocracy. Data does not deny merit. Data only shows who has more of the inputs. I fail publicly so I can learn correctly in private. I once misread a transfer because I believed a system would shape a player, when in reality the player had been dropped into a system with no consistency. The lesson was not to stop making calls. It was to never judge individual ability before checking the quality of the infrastructure that person operates inside. The same rule applies to iTero. There is no way to judge their product without a published evaluation methodology. No sample size means no confidence. No baseline means no improvement. That is the truth most AI-in-esports writing avoids: a tool can change a workflow without ever proving it improves outcomes. Those are different things, and only one is measurable on a scoreboard. Notably, the interview itself, through its two section headings, draws the line between a commercial frame and an integrity frame. The commercial frame asks who gets to use the tool. The integrity frame asks who is trusted not to abuse it. Missing is a third frame — league fairness — asking a simpler question: if this tool genuinely creates an edge, is one permanent member of a closed league holding it exclusively acceptable? I am not a prophet. I only read probability faster than you read emotion. And the probability here leans clearly: organisers will not act until a team loses an important series and blames tooling asymmetry. The accusation comes first, the rule follows. My verifiable prediction: within 18 months, at least one regional franchised league will issue a mandatory transparency clause for third-party analytics tools, requiring disclosure of input data scope and the timing at which tools may access data. A second, easier-to-check prediction: at least one team will formally complain about another team's exclusive analytics arrangement, that complaint will be dismissed on technical grounds, and it will still force the league to publish an explanatory ruling. If both predictions are wrong, I will write that here, publicly. That is the only way a hot-take writer keeps the right to write hot takes. What I want you to carry away is not a conclusion about iTero or GIANTX. It is a way of asking. Next time you see a new tool signed exclusively with a team, do not ask how smart the tool is. Ask who does not have it, and which league is letting that difference survive into a second consecutive year.