By the second coffee on day one, the stands start to blur.

This is not a criticism of anybody exhibiting. It is what happens when a category settles on a shared vocabulary faster than it settles on shared definitions. SBC Summit brings more than 800 exhibitors and 600 speakers to Lisbon across three days and six themed conference stages. Most of those companies genuinely do different things. Almost all of them will reach for the same handful of words, because those are the words that currently signal seriousness.

This piece is written for the marketing and product teams at licensed operators who will spend those three days trying to tell the difference. We will be there too, saying some of the same words. So treat what follows as a set of questions we would be happy to be asked, including the two we would answer least comfortably.

Why everything sounds the same

Look at how the agenda is organised this year and you can see the vocabulary forming in real time. The Technology and Innovation stage runs Sports Betting Innovation, iGaming Product and Design, and Marketing and Customer Experience. There are Tech Academies on AI, Web3, marketing and gamification. The Affiliate Leaders Summit moves through omnichannel marketing tech and adtech analytics.

These are the right topics. They are also the topics every vendor has spent the last twelve months learning to speak about, which means the words themselves have stopped carrying information. When personalisation can describe a rules engine, a recommendation model, or a dropdown that swaps a banner by country, the word is no longer telling you anything about what you would be buying.

So the useful work on the floor is not collecting claims. It is finding the question that makes each claim specific.

Here are the three you will hear most, and what to ask after each one.

Claim one, we personalise

Almost everybody will say this. The question that separates them is not how good the personalisation is. It is where the data comes from and where it goes.

Ask it in two parts.

Whose segments are these? If the answer is that the vendor builds the segments, you are being offered a second segmentation system alongside the one your CRM team already runs, and you should find out early whether those two will agree with each other. If the answer is that the vendor renders against segments you supply, that is a smaller, cleaner integration, and it means the work your team has already done keeps its value.

Then ask what data leaves your building. This matters more in this industry than in most, because your security and compliance review is going to ask the same thing in eight weeks and it is cheaper to know now. Ask specifically whether personal or transactional player data is stored on the vendor's side, and ask for the answer in writing rather than as a nod at a stand.

A good sign is a vendor who answers this without checking with a colleague. It means the question comes up often enough that they have a settled answer.

Claim two, marketing can run it without engineering

This one is close to a category standard now, and it is usually true in a narrow sense and misleading in a wide one.

The narrow sense is that once something is installed, a marketer can change what appears inside it from a dashboard. That is real and it is genuinely valuable, because it moves campaign changes out of the release cycle.

The wide sense, the one that gets implied, is that nothing ever needs engineering again. That is never true. The honest version has a boundary, and the useful question is where the boundary sits.

Ask what specifically still requires a developer and a release. New placement in a part of the app that has not had one before? A change to the shape of the surface rather than the content inside it? A new format that was not part of the original integration? Get them to name three things that would still put you back in the sprint queue.

Our own honest answer, since we are asking you to demand one from everybody, is that the first integration is a release. There is no version of this where a third party appears inside your app without your engineers being involved once. After that, placements can carry different widgets without further app or website code, which is the part that actually changes how a content team works week to week. But the first one is a real piece of work and anyone telling you otherwise is selling you a surprise for later.

Claim three, we use AI

This is the one where the gap between the floor and the deployment is widest, and it is not specific to this industry.

McKinsey's 2026 State of AI survey, fielded between 4 May and 8 June 2026 with 1,719 respondents across 97 countries, found that 40% of large organisations now report scaling AI agents, up from 27% the year before. At smaller organisations the figure was flat at 22%. So scaling is real, it is moving fast at the top end, and it is still a minority position. Most of what gets described in the present tense on a trade show floor sits somewhere between a pilot and that minority.

So the question is not what the AI does. It is a question about time and names.

Is it in production, with a named customer, and how long has it been running? Not in a pilot, not in beta with a friendly account, in production. If the answer is that it is coming, that is a perfectly reasonable answer and you should write down the date you were given so you can check it at ICE in January.

The second half of the question is what happens when it is wrong. Anything that generates or selects content in a regulated environment is going to be wrong occasionally. Ask who reviews the output before a player sees it, and whether the operator can turn the automation off for a specific market or campaign without turning off the product. In a licensed business, the ability to intervene is worth more than the ability to automate.

This is a question we would not enjoy being asked loosely, because parts of our own AI story are further along than others. The fair thing is for us to name which is which when you ask, and for you to hold every stand to the same standard.

The claim nobody makes, and the one that decides your timeline

There is a fourth thing worth asking about, and you will not hear anybody volunteer it, because it is the least exciting sentence at any stand.

How long does your security review take?

In a licensed business this is frequently the thing that decides whether a project lands this year or next. A vendor can be perfect on product and still cost you a quarter because their documentation pack arrives in pieces, their pentest report is eighteen months old, or nobody on their side has been through an operator's third-party risk process before.

Ask what certifications they hold and when they were last audited. Ask whether they can send a completed security questionnaire before you ask for one. Ask how many operators of your size have already put them through this, because the second time a vendor does it is always faster than the first.

The answer tells you something the demo cannot, which is whether the company has sold into your kind of business before or is about to learn on you.

The last question, and the one people forget

The four above are about the product and the process of buying it. This one is about what happens after you buy it, and in practice it predicts more of the outcome than the feature comparison does.

What does month four look like?

Month one is implementation and everybody is attentive. Month four is when the novelty has gone, the person who championed it has three other projects, and the thing either has an owner on your side or it quietly stops being used. Most tools that disappoint did not fail technically. They ran out of someone to run them.

Ask the vendor who on your team they expect to be in the platform every week, and how many hours they think it takes. A vendor who has watched this go wrong a few times will give you a specific and slightly unflattering answer. A vendor who says it runs itself has either not been doing this long or is not telling you the awkward part.

Then ask yourself the same question honestly, before you sign anything. If nobody on your side owns the weekly publishing, the platform is not the problem you need to solve first.

What to take with you

You do not need a scorecard. You need five questions and somewhere to write the answers down.

Whose segments, and what data leaves the building. What still needs a release. Is the AI in production, with whom, since when, and what happens when it is wrong. How long does the security review take and have they done it before. And who on our side runs this in month four.

Ask all five at every stand, including ours. The point is not to catch anybody out. It is that the answers vary enormously, and the variation is invisible if you only collect the claims.

One last thing worth saying out loud. The stands that answer these well are not always the ones with the best product. Sometimes they are just the ones who have been asked before. That is still useful information, because a vendor who has had these conversations a hundred times has usually also had the implementation go wrong a few times and learned something from it.

We are at SBC Summit in Lisbon from 29 September to 1 October, across Feira Internacional de Lisboa and the MEO Arena. If you want to test the five questions on somebody, we are a reasonable place to start, and we will tell you where our answers are weaker.