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

There is a seven-minute window that almost nobody in esports bothers to count correctly. It is the break between Game 2 and Game 3 of a BO5. Traditionally, those seven minutes were just enough for a coach to scribble a few lines on a whiteboard and steady the team's nerves. But in the past few seasons, that window has become the most decisive zone of a professional match — and the place where the rules lag furthest behind the technology. I have spent weeks reviewing coaching screen recordings, analysis streams, and the technology contracts nobody mentions in the industry. The more I looked, the more a paradox emerged: while fans argue over who plays better, tournament organisers argue over an entirely different question — whether an AI tool is allowed to sit with the team. That is where the name Jack Williams appears, alongside two other entities: iTero and GIANTX. Three names, one question. And that question is not about technology. It is about power. Data does not lie — only the listener lacks patience. To understand why a coaching tool becomes a governance issue, we must revisit the short history of data analytics in esports. In the early stage, around the start of the 2010s, professional teams worked with manual spreadsheets. A coach logged tower timings, objective timings, and a few crude metrics like KDA. This was the era I call analysis by memory — meaning conclusions depended on who remembered more, not who measured better. The second stage arrived when data platforms automated collection. Official publisher APIs made it possible to pull thousands of matches and millions of events. Analysts began speaking in percentages, in resource gaps by timestamp, in combat performance normalised by role. This is the rest of the scoreboard I always look at — the columns most people never scroll to, yet which decide outcomes long before the match ends. The third stage is the one now unfolding: machine learning and AI. The key difference does not lie in AI computing faster than humans. It lies in AI finding correlation patterns that the human eye misses, then proposing adjustments during the break between games. That capacity to propose is what creates the controversy. Once a tool no longer merely describes the past but suggests the future, it crosses the line between analytics and coaching. The crowd watches the score. I watch the rest of the scoreboard. iTero sits in this third stage. From what the interview with Jack Williams covers, iTero is an AI-assisted coaching platform, and the piece revolves around two larger blocks: first, the exclusive partnership with GIANTX and the likelihood of being copied; second, the risk of the tool being used for machine-assisted cheating. With just those two section headings, the picture is clear enough to reconstruct. GIANTX is an esports organisation headquartered in the EMEA region, formed from the merger of two established organisations, and tied to the regional-tier League of Legends ecosystem. This is a closed-league system — a franchised model where member teams are permanent, with no relegation. That detail matters far more than its surface appearance. In a closed league, a structural advantage is not competed away across seasons. It persists. It compounds. It becomes part of the hierarchy. Picture the mechanism. If iTero gives GIANTX a tool that analyses faster, suggests better, and shortens decision time between games, that advantage does not vanish after one loss. It stays in the system. In a league with relegation, a temporary advantage gets copied by rivals or eroded by the pressure of survival. In a closed league, no mechanism forces that advantage to level out. This is the point most people miss when discussing esports: they look at a match, while the problem lives in an entire system. On the likelihood of being copied, the story is more commercial than technical. Any technology advantage, once it appears on broadcast, becomes a target. But the gap between seeing a tool and rebuilding it is vast. Outsiders can see the output — a strange draft ban, an unusual rotation — but not the architecture beneath: which input data was chosen, which model was trained, which assumptions were discarded. Copying the surface is easy. Copying the system is nearly impossible without the right data source and the right builder. A single number is an accident. A cluster of numbers is a confession. The second block — the risk of AI-assisted cheating — is the darkest part and must be stated plainly. Two zones need to be distinguished. The first is real-time assistance during play: any external intervention in the live decision process is already clearly banned in every major title. There is nothing to debate there. The second zone — the interesting grey area — is the break between games in a BO3 or BO5. In that break, the boundary blurs. Coaches are allowed to communicate with the team. But if a coach uses an AI tool to deliver pre-computed instructions, who is making the decision? The human or the machine? And if it is the machine, is it still a game between two human teams? This is not idle philosophy. It is the question league rule-makers will be forced to answer within a few seasons. I once witnessed a similar case in football analytics: when the xG metric became widespread, people argued about whether it would ruin the natural feel of football. In the end, what changed was not football, but how people evaluated it. With esports, I expect the same script, but at far greater speed — because esports patch cycles are weekly, not annual. Here lies a variable almost nobody mentions in the AI-coaching debate: the update cadence of each title. In a title with large but infrequent updates — for example Dota 2, with its long systemic patch cycles — a machine-learning model trained on historical data retains value for longer windows. This favours statistical and ML tooling. Conversely, in a title patched every two weeks, the half-life of any learned pattern is far shorter. The implication is clear: if an AI product is marketed identically across both title types, that is a red flag. The true value of a tool inverts with the environment. Where updates are sparse, value lies in the depth of historical modelling. Where updates are dense, value lies in the speed of detecting meta shifts — a tempo advantage, not a knowledge advantage. A single product cannot optimise for both without trade-offs. This is where I must be explicit about the limits of this article itself. My source material contains no data on patches, win rates, drafts, or specific tournament structures. It holds only three content signals: the interview topic, the exclusive-partnership section, and the AI-assisted-cheating section. Every quantitative conclusion above is an inference from industry structure, not a citation from source data. I say this not to hedge, but to hold the professional line: I do not write to be agreed with. I write to be verified. One small detail is worth pausing on. In the article's biography section, there is a mention of Natus Vincere lifting the Aegis of Champions at a gamescom event fourteen years ago. If that year was 2026, simple subtraction places the article around 2026. That is not a meaningless number. It shows we are at the exact moment when a generation of esports people — those who witnessed the first era — begins shifting into tool-building roles. And when the tool-builder is someone who understands the game, the line between tool and player becomes ever harder to draw. Crisis does not create phenomena. It only exposes forgotten data. Here I want to isolate a counter-intuitive angle — the part I believe is least analysed in the whole story. The article's two section headings reflect two frames: a commercial frame (exclusivity and being copied) and an integrity frame (AI-assisted cheating). But between them sits a third frame nobody names: the league-fairness frame. The fairness question is not whether the tool is legal. It is: if only one member of a closed league can use it, where does that place the rest of the league? This is not a moral matter. It is a structural one. A competitive-preparation advantage, if not shared and not competed away, becomes a permanent advantage. And when it is permanent, it stops being a tool. It becomes part of the institution. In a relegation system, permanent advantage is resisted by survival itself: weaker teams will do anything to copy or replace it. In a closed system, that pressure does not exist. This is why I argue exclusive tooling deals are structurally more consequential in closed leagues than in open-circuit systems. And the rule-makers — the publisher or the league operator — will have to choose: either mandate shared access, or restrict the tool. History shows they have done the latter repeatedly, gradually tightening rules on in-game coach communication. There is a trap I want to flag, and it is aimed at those over-enthusiastic about AI coaching: correlation is not causation. A team using an AI tool and achieving good results does not prove the tool produced those results. The team may simply have been stronger. It may have invested in analytics overall, with AI a small part. The good results may have come from a lucky draft. To prove a tool's true value, you need a large match sample, a control group, and a method that isolates the effect from the reputation of the team using it. No platform can make a performance claim without those three. This is why I remain wary of every published performance figure: they rarely come with sample size, method, or definition. In my trade, a performance claim without method is just marketing dressed in numbers. The scoreboard does not lie, but the people reading it might. So where is the lesson in the iTero and GIANTX story? It is this: in every tooling revolution, people fixate on the tool, while the real impact lives in the rules that follow. When a tool is powerful enough to shape outcomes, the question is no longer how well it works, but who is allowed to use it. And the answer to that second question will decide the structure of an entire field for the coming decade. I once wrote that the transfer window is a chess game where most people only see the pawns. The analytics-tool story is the same. People see a contract. I see a redistribution of competitive advantage. People see a utility. I see an unwritten governance layer. What is striking is that the insiders are themselves struggling. Jack Williams, speaking about being copied, raises a question about time: how long an exclusive advantage lasts before the market erodes it. That is the right question, but not a sufficient one. The fuller question is: when that advantage erodes, what remains? If the answer is nothing, the product is not a system — it is a temporary trick. And temporary tricks do not build an industry. I expect that within 12 to 24 months, at least one major league will issue explicit rules on AI tools in competition. Not because they want to, but because they will be forced to, once exclusive deals start producing inexplicable outcomes. And when that rule arrives, it will also shape the business model of platforms like iTero — pushing them from selling advantage to selling transparency. Before you curse a coach for a strange draft ban, check your own database first. The question left for the next round is simple but hard to answer: if tomorrow an AI tool suggests the exact move that wins you the game, is the winner you, or the person who trained the tool? Football and esports do not lack stories to tell — they lack people willing to count again from the start.