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

The seven minutes between game two and game three of a best-of-five is something I have watched hundreds of times, but never has it felt this heavy.

In Dota 2 or League of Legends, that is the window where the head coach steps out of the playing room, stands in the corridor, a sheet of paper in hand carrying numbers pushed over from the analyst room. On the arena screen, the score sits at 1-1. In the headset, a player is talking about mid lane. Somewhere in the team's folder, a machine-learning model has just returned a line: if the opponent bans this champion, win probability drops six percentage points.

Back in 2026, when I was working on the organising side of a small esports tournament in Incheon, I did that job by hand. One spreadsheet, ten teams, four group-stage rounds; I typed in every metric and drew my own conclusions. Seven years later, a company called iTero is commercialising exactly what I used to do, but at industrial scale, and it has just signed an exclusivity deal with GIANTX.

The conversation with Jack Williams, the man behind iTero, does not open with technology. It opens with a far older question: when a tool can change match outcomes, who is allowed to use it, and at what point does using it get called cheating.

Context: one exclusivity deal and two headings

Jack Williams is not a name that LEC viewers or Dota 2 viewers recognise immediately. He belongs to the class of people who rarely appear on broadcast: the tool seller. iTero is his product, an analytics platform built around match data to support coaching. GIANTX is the client, and judging by how the interview was framed, the exclusive client.

One thing must be stated plainly to avoid misreading: what has been published from that interview is technically very thin. There is no sample size, no evaluation methodology, no quantitative evidence that the tool actually creates an advantage. The two headings that were disclosed, one about working exclusively with GIANTX and the likelihood of being copied, one about AI-assisted cheating, show that the story's centre of gravity sits in commerce and integrity, not in algorithms.

That in itself is a detail worth recording. In an industry where every technical advance is usually advertised with a chart, a company building AI for esports is selling its story through two legal keywords.

On organisational context: GIANTX is widely reported in industry press as an EMEA-based organisation operating in the LEC ecosystem, formed through the merger of two established organisations. I flag the phrase "widely reported" because this detail is background knowledge requiring verification and does not appear in the interview itself. But if accurate, the governing framework for the iTero and GIANTX arrangement is the publisher's rules on third-party software and competitive integrity, not the sports law of any country.

And this is where everything starts to carry weight. The LEC is a closed league. There is no relegation. Members are permanent members. In a system like that, structural advantages are not competed away across seasons. They accumulate.

One timing detail also needs anchoring. The article mentions Natus Vincere lifting the Aegis of Champions at gamescom, corresponding to The International 2026 in Dota 2. The phrase "14 years ago" in the text itself places the interview around 2026. That is arithmetic inference from the article's own words, not information stated directly.

Analysis: a tool's value inverts between two titles

The easiest measurable starting point is patch cadence.

Dota 2 runs on Valve's rhythm: large, infrequent, systemic updates, with long stretches of stability in between. League of Legends runs on Riot's rhythm: patches every two weeks, continuous, in small pieces.

Those two rhythms create two different markets for the same AI product.

In Dota 2, a model trained on historical data retains its value longer. Patches come and go, but the game's structure changes more slowly than the rate at which a model becomes obsolete. The tool's value lies in the depth of historical modelling.

In League of Legends, the two-week cycle shortens the half-life of any learned pattern. The tool's value shifts from solving the meta to detecting the meta delta faster than opponents. That is a tempo advantage, not a knowledge advantage.

A tool marketed identically across both titles is a red flag, because its core value inverts between the two environments.

The interview, at its published length and depth, does not touch this. It does not say which title iTero serves, does not say where the input data comes from, does not say how often the model is retrained. Those are not trivial technical details. They are the entire commercial story.

Selling an AI tool for esports means selling three things at once: data access, inference speed, and legality. Remove one and the product collapses.

Data access is the hardest. Professional match data sits largely in the hands of publishers or third-party aggregator platforms. A company like iTero must either buy it, scrape it, or collect it manually. Each path carries its own risk, and all three depend on whether the publisher tightens the rules.

Inference speed is the easiest to sell, because it lives inside the seven-minute window between games. There, nobody has time to run a fresh model. There, only pre-prepared results or near-instant returns have value. This is the most fragile boundary between preparation support and in-game support, and it is also where every cheating dispute will play out.

Legality is the most expensive. It cannot be bought with money, only with time and the approval of the third party that holds the rights.

Among those three, the exclusivity deal with GIANTX tells a very specific story. Exclusivity, by definition, is the limiting of how many people may use a resource. In professional esports, nearly every preparation advantage has at one point been regulated to some degree: practice hours, server access, how many coaches may speak in the headset. Analytics tools have not yet been regulated at the same level.

The central question of this decade is not whether AI helps a team win, but whether a team may be allowed to hold exclusive rights to the tool that helps it win.

And that question only truly bites inside a closed league. In a system with relegation, a disadvantage is competed away: weak teams drop, strong teams rise, the level flattens itself. In the LEC, there is no such mechanism. Members stay. If one member holds exclusive access to a tool across multiple seasons, that is no longer a temporary advantage. That is structure.

One more context note is worth adding. History shows publishers tend to regulate more tightly in proportion to impact on outcomes. In-game coach communication was once limited by wave, then adjusted further. Anything capable of affecting a game directly while it is being played gets the closest scrutiny. By contrast, things that only affect the period before the match begins tend to remain more free.

That is precisely why most of the debate about AI coaching actually sits in the grey zone between games in a best-of-three or best-of-five. Nobody argues about direct in-game assistance, because it is already clearly banned in every major title, leaving nothing to argue about. The remaining contested ground is the between-games window, where time is measured in minutes and nobody has clearly defined where assistance begins.

No performance claim from iTero can be independently verified from what has been published, because no data, sample size or evaluation method has been stated. That does not mean the product is weak. It only means that in esports we are accepting a far lower evidentiary standard than in any other industry with a comparable impact on competitive outcomes.

On the likelihood of being copied, which the article's second heading raises: it is a real concern, but I believe it is aimed at the wrong place.

In the analytics tool market, the barrier to entry rarely lies in the algorithm. The algorithm is the easiest part to copy. The barrier lies in data access and in client relationships, two things exclusivity tightens rather than loosens. If a rival wanted to copy iTero, they could copy the model portion; they could not copy the GIANTX contract.

But for the same reason, exclusivity is not a long-term shield. In a closed league, every organisation has enough money to build an internal tool if it sees a rival gaining an edge. Exclusivity retains value only until the rest of the league decides it needs an equivalent. At that point, pressure moves up to the publisher.

On AI-assisted cheating, the remaining heading: I think most of the debate will go nowhere because it demands a definition nobody has wanted to write down, namely the exact, second-by-second boundary between pre-match preparation and in-game assistance. The between-games window is neither preparation nor competition. It is territory with no law.

And that is why most AI coaching products will live in exactly that territory, sold to exactly those teams with money, inside exactly those closed leagues without relegation. The incentive architecture is pushing everything in that direction.

When a player is suspected of using an out-of-rule tool, the world names a fault, the experts name a fault, and I want to name the pressure that pushed them there. People call it a mistake; I call it a wound trying to speak. But that empathy has to have limits: empathising with pressure does not mean erasing responsibility. A banned tool is a banned tool, regardless of what the person using it was going through.

Contrarian angle: the blind spot of collective memory

The biggest blind spot in esports' collective memory is the belief that the discipline operates as a pure meritocracy. That whoever is better simply wins. For more than a decade that image was partly built by the organisations themselves: long practice sessions, closed scrims, reflex-curve numbers. All of it evokes an arena where outcomes are decided by a player's hands and brain.

Reality is more complicated. A significant share of the gap between top teams and the rest does not lie in mechanical skill but in the quality of the analytics infrastructure. We have known that for a long time about well-funded organisations: they hire more analysts, keep their own data rooms, and run more rigorous opponent assessment processes. AI is simply the next step in that same logic, repackaged and sold outward.

If that is true, then the question of whether AI constitutes cheating is mis-framed from the start. The question worth asking is who can afford to buy an advantage, and whether the league accepts letting that gap widen.

Here I want to be direct about how I choose subjects: I do not write about esports as a miniature version of football. I write about it as a market for mental labour, where exclusivity contracts and data access matter as much as reflexes. If exclusive tooling ever becomes the norm, what erodes is not the entertainment value but the story of opportunity. And that is more worrying than any single cheating accusation.

The fear of being copied, in the end, is a good sign for the industry. It means the market is thick enough to be worth copying. A company worried about being copied is a company that believes its product has value. That is the worry of a winner, not of a latecomer.

One further point rarely raised: in this ecosystem, most of the influence does not sit with the tool company but with the publisher. The publisher controls the data, controls the rules, and controls whether an exclusivity arrangement is considered valid at all. Any debate about the future of AI coaching that ignores this half is a half debate.

Takeaway

When I was growing up in the stands of a stadium in Incheon, I learned to see a match through the people sitting beside me. A boy crying, a scarf already worn thin, a row of people quietly walking out in the rain. Ten years later, I sit watching numbers run across an analyst room screen, and I wonder what audiences will remember ten years from now. Will they remember a seven-minute window, or only some moment that no tool could compute.

Before I was a journalist, I was a spectator. Before I analysed, I loved.

Tactics explain the match, but they cannot explain why our hearts beat. And if there ever comes a day when every decision on the server is made by a machine before a player can think, then what remains of esports will be the thing hardest to model: the moment a human being chooses to do what nobody predicted.