Jack Williams, iTero and GIANTX: The Grey Zone of AI Coaching Is Repricing Tactical Advantage in Esports

Six Minutes Between Game Three and Game Four

Six minutes. That is roughly how long officials allow between the third and fourth game of a best-of-five. In those six minutes an esports coach has to do three things at once: read back the loss, lock two pick-ban options, and pull five players back into rhythm. No locker room. No long half-time. No whiteboard placed in the middle of a circle. Only a screen, a headset, and however much information can be pushed into a player's head before the countdown hits zero.

That is why the interview with Jack Williams about iTero, about GIANTX, and about the future of AI coaching in esports deserves to be read slower than a news brief. Not because it announces a technical breakthrough. Because it touches the exact interval every analytical tool wants to occupy: the six minutes between two games.

What the Interview Says, and What It Does Not

On the surface, the piece rests on two axes. The first is the exclusive relationship between iTero and GIANTX, with a follow-up question attached: how likely is the product to be copied. The second is AI-assisted cheating, meaning the line between a legitimate tool and an illegitimate external influence.

The rest of the factual payload belongs to the writer's biography. One detail stands out: the article mentions Natus Vincere lifting the Aegis of Champions at Gamescom 14 years ago. That milestone maps to The International 2026. The subtraction places the article somewhere around 2026. And across the entire text there is not a single line about a patch, a bracket format, a roster, or a seeding mechanism.

To put it plainly: the piece belongs to the B2B thought-leadership genre, aimed at enterprise buyers rather than tournament audiences. Its analytical value sits in three other places: the tool's business model, the exclusivity structure, and the grey zone of rules the esports industry has not yet named.

The Real Meta Is Not Inside the Game

When patch data is zero, the only readable meta layer is the meta of the preparation trade, meaning how teams solve a patch rather than what the patch solves for them.

Dota 2 and League of Legends run on two different clocks. Valve ships infrequently and hard: one large version breaks the old structure, then the ecosystem stabilises for a long stretch. Machine learning models trained on historical data therefore retain value for months. Riot ships every two weeks. There, the half-life of any learned pattern is far shorter, and the value of AI shifts from solving the meta to detecting the meta earlier than the opponent.

The key point sits here: the same product inverts in value between the two titles. In a slow-patch title the tool sells depth of historical modelling. In a fast-patch title it sells speed. A vendor marketing both with identical language is placing itself under heavy scrutiny, because those two advantages are not the same kind of thing.

Based on my experience tracking matches, this is where people misread most often. An analytical tool does not get stronger with more data. It gets stronger with the gap between when data appears and when the opponent finishes processing it. In basketball that is the gap between Team A noticing an early switch and Team B exploiting it before the defence rotates. In esports it is the gap between a pick-ban that is off-rhythm and one that has already been predicted.

Six Minutes, and the Entire Grey Zone Sits There

AI-assisted cheating is one of the two disclosed section headings. Worth noting: in every major title, real-time assistance has been clearly prohibited for years. There is not much ground left to argue there. The genuine grey zone sits in the between-game interval, precisely the six-minute window every tool wants to enter.

Before the match, analysis is free. After the match, it is freer still. The contested boundary is the stretch where the match has started but has not ended. A model reads the first three games, produces a probability for the fourth pick-ban, and the coach decides on that probability. The tool never touches a key. It touches a decision.

I worked as an esports competitor and tournament organiser before moving into media, so I know how organisers write rules. Rules prohibit behaviour, not thought. A tool that produces a number while the match is live violates no clause if the clause only addresses software running on the tournament machine. That is a textual gap, not a moral one. And textual gaps are always patched after someone has already walked through them.

Exclusivity Is a Roster Slot, Not a Feature

The other axis of the interview is the exclusive agreement between iTero and GIANTX. This is where market analysis matters more than technical analysis.

GIANTX is known as an organisation present in the EMEA League of Legends ecosystem, formed through the merger of two established organisations. If accurate, the governing framework for this arrangement is the publisher's third-party software and competitive integrity rules. In a franchised league there is no relegation, so every participant is a permanent member. A structural advantage held by one member persists across seasons rather than being competed away as it would in an open circuit.

I remember 2026, when I was writing for a new sports outlet in Busan, publishing an analysis of the Houston Rockets and the role of P.J. Tucker, jersey number 4, averaging 6.1 points and 5.6 rebounds per game. Media at the time only mined Harden and Paul. I argued Tucker was the link holding the switch-everything system sealed, and that his defensive flexibility was what carried the Rockets to the Western Conference Finals. The piece drew 2,100 shares in 48 hours.

The lesson from Tucker maps directly here. An analytical tool does not score. It does exactly what a role player does: it keeps the system from cracking. And like a role player, its value is underrated until the system collapses without it. The worker looks at numbers, the strategist looks at flow. Here, the flow is the flow of contracts, not the flow of box scores.

When the Tool Gets Copied, the Moat Is the Data Pipeline

The interview raises the question of how likely iTero is to be copied. That is the right question asked in the wrong frame of reference.

A machine learning model is not a hard-to-copy asset. Network architecture, loss functions, feature handling can all be rebuilt by a sufficiently good engineering team within months. What is hard to copy is the data pipeline: access to closed scrim data, internal draft logs, coach feedback after every game. Nobody copies a relationship.

Transfers do not buy players, they buy expectations. An exclusive tooling agreement operates the same way. It does not buy software, it buys exclusive data access for a fixed window, and it buys negotiating position with the publisher. If iTero sold the same product to multiple teams in the same league, that product would become shared infrastructure and lose all differentiating value. Exclusivity is the product.

That is also why the copying question reads more defensively than technically. Nobody fears having their model copied. They fear losing sole-vendor status.

A Number Without a Methodology Is a Red Flag

The article contains no performance metric accompanied by sample size, measurement window, or evaluation method. The reader cannot know how many pick-bans the model called correctly. The reader cannot know how many matches it was validated on, in which league, over what period.

For a tool sold to professional teams, that absence deserves flagging. The worker looks at numbers, the strategist looks at flow. A number detached from methodology is only simulated flow.

I am not saying the product is ineffective. I am saying there is no evidence yet to assert that it is effective. The distance between those two sentences is the distance between an analytics department and a marketing department.

Two Publishers, Two Sets of Rules

Valve and Riot Games have historically approached third-party data and assistance tools differently. That difference in openness creates two entirely different addressable markets for the same AI coaching vendor.

A tool that is strictly compliant in one title may be a violation in another. Compliance cost is therefore not fixed; it scales with the number of titles a vendor wants to serve. For a small company this is a market-selection problem, not an expansion problem. And market selection will determine both product architecture and pricing model.

The Counterintuitive Angle: AI Sells Tempo, Not Knowledge

Most of the AI coaching debate assumes the tool delivers a knowledge advantage. The team with AI understands the meta better than the team without.

I think that assumption fails in most live titles.

In a fast-patch title, meta knowledge does not hold. Any pattern a model finds will surface on analyst broadcasts, on player streams, on community forums within days. Knowledge is publicised at near-instant speed. The advantage is not in knowing first, but in knowing two days early, and knowing two days early is a different animal from knowing two months early.

In a slow-patch title the story inverts. A deep historical model does produce real advantage, but that advantage can also simply be bought by hiring better people. It does not come bundled with the software.

Consequence: the most durable moat for an AI coaching vendor is not the best model, but a long-term exclusivity contract plus a closed data pipeline. The model is the most replaceable component in that entire value chain. And that is why the competitive integrity problem is misframed when people argue about cheating software.

The real integrity issue is resource asymmetry inside a closed league. Cheating is the conduct of a team violating shared rules. Asymmetry is a state in which shared rules allow one team to have more than another, legally and durably. The two differ in kind, and conflating them sends rulemaking in the wrong direction.

One condition must be stated: if the league operator mandated that every team receive access to the tool on equivalent price and terms, the entire asymmetry argument above disappears. If it does not, that argument returns every transfer window under a different name.

The Variable for Next Season

The worker's role never disappears, it is just upgraded into a system. The 2026 analyst took notes by hand. The 2026 version has a model taking notes for them. The 2030 version will manage a tool layer they must themselves audit.

Three things to watch. First, how franchised leagues rewrite the definition of external assistance to cover the between-game window. Second, whether the iTero-GIANTX model gets replicated to other organisations, turning exclusive analytics into shared infrastructure and erasing the gap just created. Third, whether the vendor publishes sample size and validation methodology.

The biggest variable is the second one. An exclusive tool is only valuable while it is exclusive. The moment it becomes a common standard, the only thing left to sell is update speed. And speed, as Mbappe showed at the 2026 World Cup, is not an invention. It is a value redefined by whoever knows how to use it at the right moment.