iTero, GIANTX and the Unwritten Line of AI Coaching in Esports

In 2026, at Gamescom, Natus Vincere lifted the Aegis of Champions. Fourteen years later, an interview with Jack Williams — the man tied to iTero — opens on exactly that milestone before turning to something else entirely: whether artificial intelligence should be allowed to sit in the coaching chair in esports. In the material I have in front of me, the substance sits in just two headings: one on iTero's exclusive partnership with GIANTX and the risk of being copied, one on AI-assisted cheating. Everything else is the author's biography. Scores lie; data is the only witness I trust. GIANTX, by the industry's own account, is an organisation present in the EMEA League of Legends ecosystem, born from the merger of Excel Esports and Giants Gaming. If that holds, the governing framework for the iTero arrangement is Riot Games' third-party software and competitive-integrity rules. To be clear: that is inference from industry data, not a claim made in the interview. The central question the interview touches is narrower than it looks. Real-time in-game assistance is already banned outright across every major title, so nothing is left to argue about there. The genuine grey zone sits between games — the few minutes in which a coach re-reads data, adjusts the draft, changes tempo. If a machine-learning model compresses that process from thirty minutes of manual analysis into three, its value is not that it coaches better than a human. It is that it shortens reaction time. That is an advantage in tempo, not in knowledge. And here is where the data forces me to pull two titles apart. Dota 2 runs on Valve's cadence: infrequent but structurally disruptive major patches, with long stable stretches in between. A model trained on historical match data keeps its value longer, so statistical tooling wins on depth. League of Legends runs on Riot's cadence, a patch every few weeks. There, the half-life of any learned pattern is far shorter, and AI's value shifts from solving the meta to detecting the meta delta faster than opponents. One product, two titles, two nearly inverted value propositions. A vendor selling one message to both markets is a warning sign, not a sign of maturity. Then comes the part I consider most neglected. In a closed league built on franchising, every participant is permanent; nobody gets relegated. A team's structural advantage is not competed away over a season — it compounds. Exclusive access to an analytics tool therefore carries far more weight than it would in an open system. League operators will soon have to choose: mandate equal access, or restrict the tool. History already walked that exact road with in-game coach communication — from permitted, to limited, to banned. From Seoul, where I work with transfer data, this story has another version. Korean teams have long operated as in-house laboratories, building their own analytics rather than buying it. When an exclusive commercial product arrives, it creates two tiers of teams: those with a paid tool and those doing it by hand. In Vietnam, where analytics budgets are far thinner, that gap is wider still. That gap is precisely what the interview's data does not see. What the data does not see, in fairness, outweighs what it does. Across the entire source there is not one patch, one format, one bracket seed, one performance metric. No sample size, no evaluation methodology, no training-data window. Which means every claim that this tool makes a difference is standing in mid-air. I follow the transfer market not to catch news but to catch patterns — and the first pattern here is this: when the seller and the buyer of a tool appear in the same interview, what is being sold is not statistics, it is belief in statistics. There is a paradox I want to put on the table. If iTero genuinely produced a measurable edge, GIANTX would have no reason to discuss being copied — an edge that can be copied is no longer an edge. The fact that insiders raise the copying risk themselves suggests the moat sits in the commercial relationship, not the technology. A correlation between using AI tools and better results does not establish causation. Richer teams buy more tools; richer teams also buy better coaches and more bootcamps. Separating those variables requires a large sample and a comparative design the interview simply does not provide. So I keep my old rule: set the error threshold before making the call. On this topic I am tracking three variables. First, each publisher's policy on third-party software, where the legal line is actually drawn. Second, the gap between LAN and online competitive rules, where real enforcement capacity gets tested. Third, how many exclusive tooling deals get signed over the next twelve months. If the third grows faster than the number of teams in the league, the franchise model is quietly creating a new tier of inequality without anyone breaking a rule. As for the AI-assisted cheating section — the integrity frame is placed correctly, but placed early. You cannot ban something you cannot define. Banning outside software is checkable via the list of processes running on a tournament machine. Banning a model pre-trained on public data that then whispers a suggestion inside a coach's head is not measurable at all. Every rule coming next will have to choose between enforceability and breadth of language. I once wrote that a crisis is just a data set nobody has cleaned yet, and it holds here. The iTero–GIANTX story is not a crisis; it is an empty data column waiting to be filled. When the noise stops, the data starts singing.