bet365

The product is mature. We maintain it and fine-tune it, and every day new segments are added, for every market. The designers I joined were strong professionals. AI was not yet part of how they worked.

I came with a doctorate in digital arts, with a thesis on AI and human-computer interaction. I had been using generative tools before this initiative, and I could see where our process, with many people passing through the same steps, could be much simpler.

AI was going to reach the company sooner or later. I wanted to be among the first, and I am glad that I was.

The app is not the same everywhere. There are region-specific versions, so it differs from market to market. That means many teams, many segments and many small variations, produced every day.

The risk was a product that looked different depending on whose hands made it.

I led the designers directly and I was accountable for how this rolled out. I also had to persuade other people to let us try new technology.

The Head of Design and the Product Owner were the key sponsors. I was given the role of lead first, then leader, and the room to apply my own knowledge and ideas about AI and about other things in the company.

These came from how I work, and from my research background.

1. AI is a tool and a colleague, not a replacement. Nobody on the team saw it as a threat.
2. Data decides where AI goes. Not enthusiasm.
3. It has to read as one product. Shared skills carry the rules, so every screen follows the same standards.
4. AI does not do the whole job. It generates well. It does not understand behavior the way we do.

Before the AI examples, here is work from the real app that our team designed: a promo module, and the payment screens, which I made with the team. The promo module is original, as it runs; the payment screens are redrawn here in the latest design.

Payment is the part of the app where trust matters most, and it is not the same everywhere. In the UK the public list includes debit cards, Apple Pay, Google Pay, PayPal, Trustly and bank transfer, and credit cards are no longer accepted for deposits. The screens above are drawn for the UK, in pounds. In other markets the list differs: in the EU, for example, it includes Sofort, paysafecard and Skrill.

Even this short flow carries the care tools. Deposit Limits and Take a Break sit on the deposit screen itself, under the amount, where someone about to pay will see them.

The navigation bar appears on almost every screen, so a small unclear detail there is repeated across the whole app. That is why I spent my time on it: five tabs and one clear active state, so that Home, Sports, In-Play, My Bets and Games each feel like a place of their own.

The skills write down the steps and the rules, so the output follows one grid and one type scale. The screens and banners below were made with AI to show the kind of output I mean. They are illustrations, not real work screens.

Bet Track lets customers follow their bets in real time from the iPhone lock screen or the Dynamic Island, so they do not have to keep opening and refreshing the app to catch the key moments of a match.

It shows the status of single bets and Bet Builder bets, key match stats, and the current cash-out amount, which updates as soon as something changes on the field. For football, American football and basketball there is also a Progress Track mode: visual indicators of how specific markets are going, such as corners, dangerous attacks or shots on target.

We made this feature with AI.

AI is excellent at generating beautiful designs. It does not yet go as deep into behavior as we do, and behavior is what we fine-tune every day.

So I did not ask it to replace that work. I asked it to clear the routine, so the team had room for it. Before writing any skill, the team and I analyzed our brand personality and what we wanted to achieve with it.

The team was excited, so “AI nowhere” was never the plan. “AI everywhere” did not survive the NDAs and the data-security rules. The decision was how to do it selectively, and by what criterion.

Regulated product, sensitive data, and AI that makes mistakes: this is where a reviewer rightly asks where I drew the line.

Mechanism

Shared AI skills inside selected segments.

Risk

Data security, NDAs, regulation, hallucinations, weak quality.

Safeguard

AI only where data allows. A person checks every output.

Who decided

Management, after the sponsors and I made the case.

Evidence level

Observation, no published figures.

I saw hallucinations, weak designs, and AI that did not understand what it was asked to do. Quality was often poor. At first the team did not want to accept it, and the team and I stayed longer hours than we should have.

Developers resisted at the beginning.

Privacy policy and country-specific rules are still a source of problems. The company has many rules and decisions it has to respect, and AI has to fit inside them.

Developers were among those who resisted AI at first. Later they began working with it too, and today the whole developer team works with AI, with limits.

The system then spread to other teams, using the foundation the team and I built.

Faster, better and more unified. More people now work as if one person did, and the team became more cohesive. I spend less time explaining, because every task has an exact skill reference that the team and I built and automated. This is what I observe, and it is not measured.

Goal

Move faster and keep teams connected

Outcome

Faster, better and more unified work; less time spent explaining; spread to other teams

Baseline, period, n

Not measured. This is an observation.

Attribution

Observation across several changes at once

Guardrail metrics

Not published

The system took root in other teams, using the foundation the team and I built. With the routine handled, the team has more time for behavioral work and research.

The product still gets daily fine-tuning, and AI is still limited to the segments where data rules allow it.

For me this is the first time I apply the ideas from my doctoral research in daily work with a team, with AI as a colleague. It freed time and space for free thinking and creative ideas.

I am grateful that I was given the lead role first and then the leader role, with room to apply my own knowledge and ideas.

The screens of the app as it looks now, redrawn in the latest design at iPhone 18 Pro size: the five main tabs and the three payment screens. Each screen starts as a pencil sketch and turns into the digital design. Hover a screen to see its sketch again. Drag or swipe to browse.

This study shows leadership of a change inside a large team. It does not show measured results, internal screens, or exact numbers, because the work is confidential. The Video Wall module is original; the payment screens and the screens in the header are based on the real app. The sketches are schematics I drew. The team images, the example screens and the example banners are AI-generated illustrations, not real work screens.