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

A Night in Seoul, a Head Count, and a Name Read Wrong

Nights in Seoul usually begin with the hum of an air conditioner rather than the roar of a crowd. I sat in front of my screen, switched off the match recording I had just rewatched, and opened a long interview about AI coaching in esports. The piece promised Jack Williams, a tool called iTero, an organisation called GIANTX, and the future of a profession still forming in front of us. I read it once. Then I read it again, this time with a pencil in hand.

I counted thirteen information points. Ten of them were about the person asking the questions: a reporter named Ollie, who loves Natus Vincere, who remembers Gamescom fourteen years ago, who dreams of lifting the Aegis of Champions one day. Only three points touched the subject promised by the headline, and two of those three were section headings with no body text behind them.

I sat still for a while. Listening before commentating is how I correct my own mistakes, a lesson I gave myself in the summer of 2026, when I mispronounced a nineteen-year-old player's name three times live in front of more than thirty-one thousand people packed into Sang-am Stadium. I learned that a broadcast can get every event right and still be wrong, if the storyteller refuses to read the name in front of him.

Tonight, the thing read wrong was an article. It claims to be about AI. Most of its length belongs to its author.

What We Actually Hold

The interview centres on Jack Williams, the iTero tool, the GIANTX organisation, and the future of AI coaching. That is the entire spine the headline promised. The flesh sits in two named sections: first, iTero working exclusively with GIANTX and the likelihood of being copied; second, AI-assisted cheating. Both appear only as headings, without data, quotes, or specific dates.

And that is all. No game version is named. No patch, no balance window, no item or map change. No tournament format, no bracket, no seeding, no schedule. Not a single player, roster, champion pool, win rate, or named statistic. In an article about the future of coaching, no coach is ever named.

The only timestamped thing is a memory. The piece mentions Natus Vincere lifting the Aegis of Champions at Gamescom, with the phrase fourteen years ago. The International debuted in 2026 at Gamescom in Cologne, where Natus Vincere beat EHOME in the final and took home one million US dollars from a total prize pool of 1.6 million dollars published by Valve. Simple subtraction places the article around 2026. The only temporal anchor is the writer's nostalgia, not the interviewee's data.

On product performance, there is nothing to hold. No sample size, no evaluation method, no metric, no test window. A performance claim without methodology cannot be verified, and that holds for every analytics product in this industry, including the ones I believe are real and useful.

Based on my experience following matches, I notice that analytics vendors share one habit: they sell the final result and hide the middle process. The buyer here is an esports organisation, not a fan, which is why the habit survives. The outside reader is left with the shell.

One Letter X, and Why Names Matter

The material calls the organisation two different ways. Sometimes Giant X, sometimes GIANTX. For many readers that is a trifle. For me, it is where the article starts to loosen. Every player's name is a short poem, if we bother to read it closely, and that applies to team names, organisation names, and tool names. A name capitalised wrong in a news item tells you the writer never typed it into a search box.

According to industry accounts that still require verification, GIANTX was formed by the merger of Excel Esports and Giants Gaming, carries an EMEA identity, and competes in Riot Games' European regional system. If that holds, the governance framework around the iTero arrangement is Riot's third-party software and competitive integrity rules, not Valve's. That detail is decisive, and it is absent from the disclosed content.

I learned the weight of names in that summer of 2026, when I wrote a letter of apology to a young player. The name I read wrong back then now rings like a song, and I no longer treat correct pronunciation as politeness. It is a matter of fact. In an industry where organisations rename, merge, and rebrand several times a decade, a misspelled name is usually the first sign that a piece was written from a press release rather than from observation.

Patch Cadence Is a Commercial Variable, Not a Technical One

The source article offers no patch information whatsoever. None of the thirteen points carries a game version field. So what follows is not sourced from the article. It rests on an observation I consider the single largest commercial variable in AI coaching, and one the piece ignores entirely.

The value of an AI tool is not in the model. It is in the rate of change of the game it serves. Dota 2 follows Valve's cadence: large but infrequent systemic patches, with months-long stretches of stability between them. A model trained on historical data keeps its validity there. The reward goes to patience: deep historical modelling, a sample size that accumulates over time, an understanding of hero interactions across match phases built on thick data.

League of Legends runs a different clock. Riot patches roughly every two weeks, and each patch is small enough not to break the game but large enough to shift the balance conditions. The half-life of any learned pattern keeps shrinking. In that environment, the value of an AI tool shifts from solving the meta to detecting the meta drift faster than opponents. That is a tempo advantage, not a knowledge advantage.

These two advantages have very different lifespans. A knowledge advantage compounds and protects itself if a team keeps its people and its process. A tempo advantage evaporates within a split unless it is refuelled continuously with fresh data and with people who know which questions to ask. A product marketed with the same value story across both cadences is suspicious. The same promise cannot hold in two worlds spinning at different speeds.

One more variable goes unmentioned: the tournament server. Professional play runs on a locked version, training data has a cut-off, and in-tournament analysis draws from a different build. The gap between those two things determines which tools remain useful and which become an echo. Without patch cadence, server lock rules, and data-availability windows, any assessment of iTero's durable edge is guesswork.

Across many nights re-watching European regional matches, I noticed something small: inside a single patch, teams draft almost identically for three weeks, then suddenly diverge. Whoever decodes that divergence first wins the next two rounds. A tool that detects the divergence twelve hours earlier is worth far more than a tool that summarises what everyone already knows.

The Between-Game Window: Where the Rules Have Not Caught Up

AI-assisted cheating is the second named section. To discuss it properly, you have to slice time into windows. Real-time in-game assistance is already clearly banned in every major title. There is nothing to debate there, so a serious piece on AI cheating necessarily concerns the remaining windows.

Pre-match is a mild grey zone. Public data analysis, VOD review, opponent draft-tendency statistics: all legitimate, and teams have done this since dedicated analysts appeared. Post-match is the same. Nobody bans drawing lessons.

The window that deserves worry is the break between games in a best-of-three or best-of-five. That is when a coach walks out, the team sits for five to eight minutes, and returns with an adjusted draft plan. If a tool feeds that window, its effect is real and its legal position is murky. Current rulebooks were written for humans with headsets, not for models calling through an API. The distance between those two subjects is exactly where the cheating boundary stretches.

One thing the source never says deserves stating: the legal line runs through data provenance, not through the presence of a model. Reading an opponent's tendencies from public data is legitimate, even with AI. Reading the same tendencies from a private feed the opponent does not know is being read is not, even without AI. Most public debate misreads this, because it fixates on the word AI and skips the word source.

For a professional-facing tool, the first question is not how strong the model is, but where the input data comes from, who consented to its arrival, and whether it passes through a third party. Without those three answers, every claim of legitimacy is just belief.

Kazan taught me that tears can be a pass. But tears only become a pass when a scoreboard stands behind them, so people know what they are crying about. A tool without a scoreboard does not create a match. It creates beautiful slides.

Closed League, Closed Advantage

Assume GIANTX competes in a franchised, closed league like Riot's European system. If so, an exclusivity deal between iTero and one permanent member carries far more weight than a club buying a better training device.

In a closed system there is no relegation. Members stay across seasons. A structural advantage held by one member is not competed away by promotion and relegation turnover; it compounds through seasons, transfer windows, and academy classes. In an open circuit, advantage dilutes as new teams rise and old ones fall, as people and data keep circulating. In a franchise, advantage freezes.

The industry's own history offers a precedent. In-match coach communication was widened and then narrowed across eras, depending on how much influence it was judged to have on results. Publishers intervene when something genuinely affects outcomes. An exclusive analytics tool sits squarely in that zone.

Consequently, a league operator has two paths. One is to mandate equal access, turning professional-grade data into shared infrastructure, as replay data once was. The other is to restrict the tool, making it an object requiring disclosure and control. Both paths demand something nobody has yet: a clear definition of when a model counts as a preparation tool and when it counts as a sixth player without a jersey.

The first section, on being copied, is a purely commercial problem. The competitive moat of an analytics product rarely sits in model architecture. It sits in the data pipeline and in the exclusive relationship with the first organisation willing to adopt it. Which means the exclusivity deal is not the reward for the product. The exclusivity deal is the product. Copying a model is easy; copying access rights is not.

The empty seat said more that day than any crowd. In this whole story, that seat belongs to the team without the tool, the player never asked, and the tier-two ecosystem never named in an article about the future of coaching.

The Third Door Nobody Opened

The article opens two doors. The first is commercial: exclusivity and the risk of copying. The second is integrity: AI-assisted cheating. Between them sits a third door, and nobody opens it. That door is intra-league fairness.

This is the largest blind spot, and it is structural rather than perceptual. Commerce is the vendor's story. Integrity is the regulator's story. Fairness is the players' story, and players are the only party absent from the article. A perspective without players cannot be called complete, however correct it may be.

The second blind spot lies in collective memory. We are living through a period that narrates AI as if sports analytics had just been born. Professional teams have hired data analysts for over a decade. Natus Vincere lifted the Aegis of Champions in 2026, at a time when reviewing recordings was itself an edge. What is new is not the existence of analysis. What is new is speed and scale. When collective memory forgets its baseline, every new tool looks like an apocalypse and every rebuttal looks backward.

The third blind spot lies in the article's own form. An interview in which the interviewer occupies seventy percent of the material is a product of a corporate content economy, where the founder's origin story is the merchandise. That genre is not wrong, but it must be read as what it is: a communications document, not an investigation.

The final blind spot is people. Not one player is named. Not one coach is named. An article about the future of coaching, with no real people in it, cannot be verified and cannot resonate. I fell in love with this profession in Kazan, during the three seconds my voice broke on radio when South Korea beat Germany two-nil, Kim Young-gwon scoring in the ninetieth minute and Son Heung-min sealing it in the ninety-sixth. What I remember is not a model. What I remember is a generation.

My voice broke in Kazan, and from that moment I knew which sounds were real. A product can be described in thousands of words. A person needs only one name, read correctly.

What I Keep

The question is not whether AI should exist in esports. AI has been in the analysis room for years and is not leaving. The real question is who owns the data window, and whether a publisher treats access to it as shared infrastructure or as a product for sale. If infrastructure, we get a fairer analytical era. If a product, we get a league where a title is partly decided at a contract table rather than on the stage. And then the only thing left to verify will be names. Real names, pronounced correctly, beside a scoreboard that does not lie.