Jack Williams, iTero and GIANTX: The Grey Zone of AI Coaching in Esports

The break between game two and game three of a best-of-five lasts somewhere between 12 and 15 minutes. In that window, an esports coach can rewatch three major teamfights, adjust two draft picks, and write four notes for the next game. If an artificial intelligence tool does most of that work in 90 seconds, the central question is no longer how strong the tool is. The central question is who is allowed to use it, inside which time window, and how long before that advantage hardens into systemic unfairness.

Jack Williams's interview about iTero, GIANTX, and the future of AI coaching in esports lands exactly inside that gap. But to read it properly, I need to say something first about the quality of the source.

Football and esports differ on the surface, but the same layer of data sits underneath. Six years of logging football data before moving into esports reporting for the US market taught me that every argument about an analytics tool eventually reduces to three questions. Where does the data come from. Who gets access. And who controls that access. iTero and GIANTX are a test of all three.

The source material and its limits

Within the material I have, the article discloses only two substantive section headings. The first concerns Jack Williams working exclusively with GIANTX and the likelihood of the model being copied. The second addresses AI-assisted cheating. Everything else describes the author of the article rather than the subject of the interview.

The consequences are concrete. I have no patch data. I have no tournament format. I have no performance metric. I have no player names to analyse. And I will not invent any of them just to make the piece look fuller. That is a rule I set for myself in 2026, when I was logging more than 1,200 World Cup shots on a spreadsheet I built myself. Without clear data, I publish no line of analysis. That rule still holds now that I cover esports, where most public information is marketing.

One historical detail is worth noting so the article is not misread. The interview mentions Natus Vincere winning The International at Gamescom in August 2026, when the team lifted the Aegis of Champions. That is the author's biographical colour, not a signal about the present Dota 2 competitive landscape. Treating it as tactical evidence would be a category error on the first line.

GIANTX is widely reported by industry media as an EMEA-based organisation competing in the LEC, formed through the merger of Excel Esports and Giants Gaming. If that is accurate, the governing framework for the iTero arrangement is Riot Games' rules on third-party software and competitive integrity. This is an assumption requiring verification, not a confirmed fact. I mark it at medium confidence and will say so to anyone who asks me about this piece.

Patch cadence is a first-order commercial variable

For a tool that learns from historical data, the release cadence of the game determines the lifespan of every model.

Valve updates Dota 2 on an infrequent, heavy cadence. A few large patches a year, with long stretches of stability between them. A model trained on old match data retains its value over a longer window, which means the advantage tilts toward statistical tooling that digs deep into history. You can teach a system to read thousands of old matches without retraining every week.

Riot Games runs League of Legends on a two-week cycle. Iteration is fast, and the half-life of any learned pattern shortens with it. Here, the value of AI shifts from solving the meta to detecting the meta delta faster than opponents do.

Those two cadences produce two products that differ in kind. One sells knowledge, the other sells speed. If iTero is marketed identically across both titles, that is the first warning sign. No model is simultaneously excellent at deep historical modelling and instant reaction at the same cost. Sports analytics learned this lesson long ago: a model optimised for one competitive cycle is often useless in the next.

I raise two counterexamples to test my own reasoning, a habit I have kept for six years. Counterexample one: a team buys an AI tool and wins three matches in a row. That streak does not prove the tool caused the result. Counterexample two: an organisation uses no AI, has a better human analyst, and still wins. Both scenarios are compatible with the hypothesis that AI tooling makes no difference at all. That is why I refuse any conclusion drawn from a win streak, in football or in esports. I do not predict the future by intuition; I only read the traces the data leaves behind.

Resource asymmetry inside a closed league

The LEC is a closed league. There is no relegation, and every member is a permanent member. In that model, a structural advantage is not competed away across seasons. It accumulates. An exclusive agreement over an analytics tool sits in the same regulatory category as any other match-preparation advantage, differing only in how hard it is to see from the outside.

This places league operators before two familiar choices. Either mandate equal access for every team, or restrict the tool. The industry already walked that exact road once with in-game coach communication rules: from permitted, to time-window-limited, to fully banned. That cycle took several years. I expect the next cycle around analytics tools to take a similar shape, only faster.

What stands out is that the two disclosed section headings of the interview cover the commercial frame and the integrity frame, but leave empty the frame sitting between them: league fairness. That is the least explored angle, and the one I care about most.

The real grey zone is the between-game window

Real-time assistance while a match is live is already explicitly banned in every major title. There is nothing left to debate there. What remains debatable is the window between games in a best-of-three or best-of-five: those 12 to 15 minutes when a coach is allowed to talk to the team.

If an AI tool enters precisely that window, it does not technically violate the ban on in-game assistance. Competitively, it may be equivalent to having a sixth analyst in the room, one who never tires and never forgets. This is where current rules have no language. And based on my experience following matches, gaps in the language of a rulebook are always filled by whichever team reads the rules most carefully.

I wrote about a similar case in football, when VAR did not reduce controversy but merely moved it from the pitch to the review room. The mechanism here is identical. A new tool does not erase the grey zone. It relocates the grey zone somewhere fewer people are watching.

The moat is not in the source code

There is a counterintuitive conclusion I consider the most important in this entire subject. If iTero's worry is being copied, that worry is misplaced.

The source code of a machine learning model is not a moat. It can be reproduced in a few months with enough data and enough people. What cannot be copied is exclusive data access, relationships with league operators, and labelled datasets nobody else holds. iTero's moat, if it exists, is in the contracts rather than the algorithm.

This also sets a ceiling on any performance claim. In the material I have, there is no sample size, no evaluation methodology, no metric with a confidence interval. Every dataset is a scripture, and I am a slow reader. Reading slowly here means a tool that has not published its evaluation methodology cannot yet be treated as having proven its effectiveness, regardless of which team is using it.

And if the fast patch cadence of League of Legends makes knowledge advantages evaporate within two weeks, then a genuine AI product does not sell knowledge. It sells processing speed. Speed is the easiest thing to copy of everything a technology company can own.

What this means for a market like Vietnam

In Vietnam, professional teams largely operate on analytics budgets far smaller than LEC organisations. If AI coaching tools become standard and are sold under exclusive licences, the resource gap will no longer sit in the quality of the players. It will sit in tool access.

That is why I track agreements like iTero and GIANTX more closely than I track major matches. An exclusivity deal signed today can shape the ecosystem for three seasons, while a single win survives in the standings for a few days. I do not predict the future by guessing which team gets stronger. I read the traces that structure leaves behind before the results appear.

Takeaway

Over the next cycle, I will be watching whether LEC operators introduce regulatory language for third-party analytics tools. I will be watching whether iTero publishes an evaluation methodology with a sample size, or only publishes outcomes. And I will be watching whether any team declines the tool and still holds its position.

Those three signals answer a far bigger question than whether AI is good. They answer who is allowed to hold an advantage, and how long before that advantage becomes the default. For anyone patient enough to wait a season to prove a number.