iTero, Jack Williams and the Rulebook Gap AI Coaching Left Behind at GIANTX

Eight Minutes of Break, and What the Rulebook Cannot See

Between game two and game three of a best-of-five, the break lasts exactly eight minutes. During those eight minutes, no coach's keyboard may touch a machine. No line of code may execute while the match is still live. This is the limit every major league erected years ago, and it is also the boundary Jack Williams — the man behind iTero — must operate inside. The conversation about AI coaching in esports has therefore stopped being a technology story. It is a story about a rulebook that was never finished.

I have watched between-game analysis sessions across multiple leagues for seven years. The lesson was simple: when the rules stay silent, the advantage does not vanish — it just moves somewhere harder to see. iTero is standing exactly there.

Context: one interview, two headings, and one piece of arithmetic

Jack Williams appeared in an interview discussing iTero, GIANTX, and the future of AI-assisted coaching in esports. The piece names two content sections. First, iTero's exclusive partnership with GIANTX and the likelihood of being copied. Second, the question of AI-assisted cheating.

The date of the piece can be pinned by arithmetic. It references Natus Vincere lifting the Aegis of Champions at Gamescom "14 years ago." Na'Vi won the first The International at Gamescom in 2026. Add fourteen years and the figure lands in 2026. That number comes from addition, not from guesswork.

GIANTX is an EMEA-rooted esports organisation formed through the merger of Excel Esports and Giants Gaming, tied to the LEC ecosystem — a league that runs on a franchise model with no relegation. That detail matters more than it looks. In an open circuit, structural advantages erode on their own: weak teams drop out, strong teams rise, and the gap closes over time. In a franchise league, every participant is a permanent member. A structural advantage persists across seasons instead of being competed away. Most of the AI debate in esports skips precisely this point.

Something must be said about the source's genre. This is a B2B thought-leadership piece, not tournament reportage. That genre rarely carries operational data, and this piece is no exception. It carries headings, entity names, and a career anecdote. That is the entire raw material.

One disclaimer about sourcing is required: the interview discloses no performance data for iTero. No sample size, no evaluation method, no controlled win rate. Everything below is reasoned from product structure and entity names, not from verified numbers. That is the first thing to note, not the last.

Core analysis: three time windows and one data pipeline

Let us pull iTero away from the question "can AI replace a coach." That question is operationally meaningless. The real question is: which questions does AI answer, inside which time window, and who is allowed to hear the answer.

Every analytics tool today operates in three windows. Pre-match: opponent data, pick-ban trends, composition probabilities. Mid-series: adjustments after each game. Post-match: review and debrief. AI delivers its highest value in the first and third windows, because both allow long processing time. The second window — the eight minutes between games — is the most legally sensitive, and the one iTero most likely targets, because that is where decisions get made under real time pressure.

Do not argue with words; let the data speak. The issue is not whether AI is good. The issue is speed: whoever spots the meta shift first wins the next game.

This is where exclusivity becomes central. iTero's exclusive arrangement with GIANTX creates two opposite kinds of risk at once.

The first risk sits on GIANTX's side. If the tool genuinely creates an edge, exclusivity is an asset. If the tool only creates the feeling of an edge, exclusivity is a sunk cost dressed in technology language. There is no way to tell the two apart without a control dataset — and as stated, that dataset is absent from the source.

The second risk sits on iTero's side, and this is the most underrated part. Exclusivity is a fast sales motion, but it is also self-limitation. One exclusive client means one dataset. One dataset means a model that learns one playstyle, not an entire meta. Data diversity is what determines long-run model quality, not contract value.

The "likelihood of being copied" question therefore needs reframing. If iTero's moat is the model, it will be copied. Every model gets copied; it is only a matter of time and resources. If the moat is the data pipeline — how data is collected, cleaned, labelled and refreshed game by game — copying becomes far harder, because it demands relationships with teams, not merely engineering.

Spreadsheets do not lie; readers must learn to listen. In esports, the fastest thing to be copied is always the interface. The slowest thing to be copied is always the data pipeline.

As for AI-assisted cheating: the question was mis-framed from the start. Competitive cheating — real-time influence on a live match — has been clearly banned in every major title for years. There is nothing to debate there. The genuine grey zone sits between games. If AI pre-builds a decision tree for situation X in game three, and the coach reads it aloud during the eight-minute break, is that cheating or coaching?

Remove AI from the equation and the question remains unchanged. Is a coach who prepared a script in advance cheating? No. So how does a tool that generates that script differ from a thicker, faster, unforgetting notebook? The real boundary lies in two places: whether AI runs while the match is live, and whether a human remains the final decision-maker. Every future regulation will orbit exactly those two questions.

Notably, the two disclosed headings cover only two frames: the commercial frame (exclusivity and copying) and the integrity frame (cheating). The third frame between them — league fairness — is entirely absent. That is the largest gap in the whole discussion, and the first gap franchise leagues will have to fill.

The contrarian angle: the risk is not cheating, it is preparation inequality

What few say out loud: the biggest risk of AI coaching is not cheating, but inequality in the preparation phase inside a closed league.

In a franchise system like the LEC, no relegation mechanism forces advantages to level out. If a permanent member holds exclusive access to an analytics tool that genuinely affects results, that advantage persists season after season. The league operator will be forced to choose: mandate equal access, or restrict the tool. History has walked this exact road once already — with the rules on in-game coach communication. First a grey zone, then a concrete regulation, finally a default.

They told me not to talk tactics, so I drew charts instead. The same applies here: every AI debate in esports will end in a rulebook table, not in a blog post.

One more overlooked point: patch cadence determines AI's real value. In a slow-patching title like Dota 2, a model trained on historical data stays valid across a long window — good for statistical and machine-learning approaches. In a fortnightly-patching title like League of Legends, the lifespan of any learned pattern shrinks. AI's value shifts from "solving the meta" to "detecting the meta delta faster than opponents" — a tempo advantage, not a knowledge advantage. A product marketed identically across both kinds of titles is a warning sign, not a strength.

And here a methodological caveat is due. The entire argument above is built from product structure, entity names, and two published headings. It is not built from operational data. If iTero publishes a sample size, an evaluation method, and a controlled win rate, the conclusion could flip. Until then, every claim about the tool's effectiveness sits at the hypothesis level, and I state my confidence explicitly: medium.

Takeaway

When I forecast, I do not look at emotion, I look at PPDA. With iTero, the metric I want to see is not how good the AI is, but how many teams will be permitted to use it next season, and when the league operator writes the rule. An out-of-place number can be a truth hiding where nobody thought to look. The open question I leave behind: if GIANTX wins more next season, who will be the first to check whether that came from the humans, or from the data pipeline behind them?