Jack Williams, iTero and GIANTX: When an AI Tool Becomes One Team's Exclusive Asset

Jack Williams, iTero and GIANTX: When an AI Tool Becomes One Team's Exclusive Asset

Between game two and game three of a BO5, there are roughly twelve minutes. In those twelve minutes, the person in the coaching chair has to answer three questions: what did the opponent change in draft, is that change a variation or a system, and if it is a system, which counter has the highest win probability. Those three questions used to be answered by collective memory, by an assistant coach's notebook, and by the intuition of someone who played that title at the highest level. Now they can be answered by a model loaded with the last three seasons of the opponent's own data.

That is the anchor for the name now circulating among esports analysts: iTero. And that is why a conversation with Jack Williams — the figure attached to the exclusive partnership between iTero and GIANTX — deserves a slower read than the normal news cycle allows. Not because it carries a large dataset. Because it puts a hand exactly on the place the rulebooks have not yet written to.

Context: three names, two legal frames, one gap

Jack Williams is the central figure of the original interview, tied to iTero — an analytics and coaching platform with a machine-learning component, aimed at supporting a team's preparation and tactical adjustment. GIANTX, the named partner, is an EMEA esports organisation formed through the merger of Excel Esports and Giants Gaming, present within Riot Games' closed European league ecosystem.

The original piece contains two clearly recorded sections. The first deals with working exclusively with GIANTX and the likelihood of the model being copied. The second deals with AI-assisted cheating. Those two sections read like two different stories — a commercial one and an integrity one. They sit on the same axis, and on that axis there is a third frame both sections glide past: fairness inside the league itself.

I have to be blunt about the limits of the source. The original piece discloses no patch version, no specific tournament beyond proper names, no sample size, no methodology for evaluating the tool's performance. Any number concerning win rate, pick-ban rate, or metric improvement does not exist in the source material. For someone who reads data for a living, that gap is the single most important fact in the whole story — not an omission, but the information itself.

One detail is worth noting for dating. The original mentions Natus Vincere lifting the Aegis of Champions at gamescom, fourteen years before the writing. Na'Vi won the first The International in 2026, held at gamescom in Cologne. The subtraction gives a date of roughly 2026. That is a verifiable arithmetic fact, and it matters because it places the AI-coaching debate exactly in the period when publishers began tightening rules on third-party tooling.

One distinction must be kept clean: the Na'Vi and Aegis detail appears as biographical colour, not as analytical data about the current competitive window. Dota 2 is mentioned as a memory, not as a tactical front under examination. Reading it as a signal about today's Dota 2 landscape is a category error. I flag this because in my profession, category errors are the most expensive kind: they do not corrupt a single number, they corrupt the entire chain of reasoning behind that number.

Core analysis: patch cadence decides a model's value

There is no patch in the source material, so the only meta layer that can be analysed honestly is the meta of the preparation tooling itself — that is, how teams solve the patch is changing.

A structural difference between titles matters here, and it is a first-order commercial variable for any AI-coaching vendor.

In Dota 2, Valve operates on a cadence of large, infrequent, system-breaking updates, with long stable stretches between them. In that environment, a model trained on historical data retains validity over a longer window. The advantage tilts toward deep historical modelling.

In League of Legends, Riot Games operates on a two-week patch cadence. That cadence shortens the half-life of any learned pattern. Here the value of an AI tool shifts from solving the meta to detecting the meta delta faster than opponents. That is a tempo advantage, not a knowledge advantage. The two have entirely different business models: a knowledge advantage sells as a long-term licence, a tempo advantage only sells as a cyclical subscription.

Three moments in my own career forced me to write differently from my instinct, and all three connect directly to today's story.

The first was the 2026 World Cup. I was fifteen, writing a blog built on expected goals, arguing against the view that Croatia had simply been lucky in the semi-final. I rewatched all seven Croatia matches, minute by minute, to answer with a chain of raw data rather than emotional debate. The piece was mocked for a child lecturing experts. The lesson I kept was not about being right. It was that a conclusion without raw data behind it is just an opinion presented more loudly.

The second was the 2026 World Cup, Morocco eliminating Spain in the round of sixteen. Commentators called it a miracle. Morocco's PPDA was 8.2 — meaning they pressed aggressively high up the pitch, with no negative defending at all. What was called a miracle was a tightly organised pressing system. I dropped the words lucky and surprising from my analytical vocabulary after that.

The third was Euro 2026. I calculated that Jamal Musiala was running about eight percent more than his own baseline and predicted he would run out of fuel in the quarter-final. The prediction held. But an editor told me to my face that I wrote like a machine and that fans hated it. I argued, then accepted he was half right. I kept the data conclusion but changed the delivery: start with the person, then bring the numbers in.

Those three moments taught me something that applies directly to the iTero story: when a tool claims to improve performance, the first question is not how smart the tool is, but what time window it operates in, at what rate of change, and who can verify the result.

Applying that frame to iTero surfaces three layers.

The first is the model's validity period. A tool that learns from history is only useful while the patterns in that history are alive. A title's patch cadence determines that lifespan. A tool designed for a slow-patch title loses value quickly if transplanted unchanged into a fast-patch title, and vice versa. If a single product is marketed identically across both, that is a flag worth checking.

The second is the application window. Real-time in-game assistance is already unambiguously prohibited in every major title. The remaining grey zone is the between-game window in a BO3 or BO5, and the pre-match preparation phase. That is why this interview is worth reading: it is not arguing about a zone the rules have closed, it is arguing about a zone the rules have not touched.

The third is the commercial relationship. An exclusivity arrangement between a tooling vendor and one member of a closed league raises a question about the symmetry of the playing field. In a franchised league, members are permanent, with no relegation pressure. A structural advantage held by one member is not competed away across seasons. It accumulates.

Here I must state the methodological limit clearly. I have no sample size, no control group, no performance data for iTero. With n equal to zero, no quantitative conclusion is permitted. What I have is a structural argument, and it is valid only as a set of questions to verify, not as a verdict.

One market context belongs beside this story. The current cycle is the transfer window. Transfer-window noise always drowns signal, and in esports the noise is thicker because contract lifecycles are short and roster value moves fast.

An exclusive coaching tool enters that market in a way few have priced. It does not directly raise a player's price. It raises the value of retaining a player who is already good, because the tool multiplies the output of someone who can already read the game. Conversely, on a young, inexperienced roster, the same tool may produce more data without producing more correct decisions. The tool amplifies existing differences more than it flattens them. So an exclusive tooling deal is, in the long run, a transfer-structure deal rather than merely a technical one.

There is a useful historical parallel. When leagues began restricting what coaches could say during a live match, they did so not because coaches had broken a rule, but because the existence of that communication channel created an unevenly distributed advantage. The logic regulators now apply to AI tooling has the same structure. A tool does not have to break a rule to become a problem. It only has to create an asymmetric advantage.

There are two paths a league operator can take when that pressure appears. One is to mandate equal access, turning the tool from private asset into shared league infrastructure. The other is to restrict the tool itself, turning it into a regulated category. Both reduce the economic value of an exclusivity deal, through different mechanisms. That is why the section on being copied should not be read as an intellectual-property worry, but as a calculation of the window before the rules change.

Contrarian angle: the missing frame is league fairness, not integrity

The two sections recorded in the original are exclusivity and copying, and AI-assisted cheating. Both are familiar frames, easy to package as headlines. The third frame sits between them and is almost never named: the internal fairness of a closed league when one member holds private access to a tool that affects competitive outcomes.

The contrarian point is this: copying is not the biggest threat to an exclusive tooling deal. Legitimisation is. An opponent who copies your product still leaves you a market, customers, and a brand story. But a publisher declaring that this class of tool must be offered equally to all teams in the league, or must have its functionality restricted, deletes not just the competitive edge but the revenue model. The barrier for a copier is low. The barrier of regulation is close to absolute.

A warning about correlation and causation also belongs here. If over the next few seasons teams using AI tools win more, that does not prove the tools made them win. Teams with stronger finances are usually the first to sign exclusive technology deals, and are usually the teams with better rosters. Two variables moving together does not mean one causes the other. Separating them requires control-level data: teams of comparable strength differing in tool access, tracked across multiple seasons, with a stated sample size. That data does not exist publicly today.

The reverse question should be asked too, to avoid one-sided criticism: if this tool is genuinely effective and that effectiveness persists, the exclusivity advantage will be eroded by the market itself, because other teams will pay for an equivalent product or build internal capability. In that scenario, the exclusive deal is a temporary, time-limited edge, and its real value lies in selling to the next team. That is entirely possible. I do not rule it out. I only say that asserting it requires data the current source does not provide.

One further point belongs to the development pathway. In recent years many former pros have opened youth academies, and most operate as commercial channels rather than systematic training institutions. If AI tooling becomes standard at the professional preparation level, the biggest gap is not at that level. It is at the grassroots coaching tier, where almost nobody is trained to read data. A good tool placed in the hands of an untrained grassroots coach creates the feeling of professionalism without the substance of expertise. That gap is not closed by selling more software licences.

Takeaway

The iTero and GIANTX story can be told as a technology item, and if it is, it will wash past within a week. What makes it worth keeping is a structural question: when a tool capable of affecting competitive outcomes is signed exclusively to one member of a closed league, who is responsible for the symmetry of the playing field.

Three signals I will track in the coming cycles.

First, regulatory signals from the publisher. If Riot Games or Valve issues any adjustment on third-party tooling during preparation and between-game windows, that indicates league-fairness pressure has moved from internal meetings into written policy. This is the highest-confidence and easiest-to-verify signal.

Second, signals from teams not granted access. If a team in the same league publicly demands equal access, the exclusivity arrangement moves from asset status to dispute status. Historically, that is often the step before institutional change.

Third, product-quality signals. If iTero or its competitors publish evaluation methodology, sample size, and applicable time windows, that marks the maturation of an entire market segment. Publishing methodology is stronger evidence than any single win-rate figure.

At twenty-three, I learned that a team does not lack stars — it lacks someone who can read the flow of the match. In esports that sentence is true in another sense: the industry does not lack data-reading tools. It lacks someone willing to answer who is allowed to read that same stream of data.