Three Gaming and Mobile Experts on Onboarding, AI, and Knowing When to Stay Quiet

AI filtering is increasingly sitting between the send button and the notification tray. Whether it’s  on iOS or in Gmail, mobile teams are finding that AI-mediated operating systems are starting to decide what gets surfaced and what disappears. Our 2026 State of Customer Engagement Report found that 48% of teams are already monitoring its impact on reach, and 17% say it's actively affecting deliverability right now.

Behavior-triggered messages are outperforming standard sends by up to 9x on CTR. Teams using automated cross-channel journeys are getting better results while sending fewer messages.

We brought in three mobile engagement experts practitioners and asked them to pressure-test the findings against what they're actually seeing:

• Baris Girgin, Associate Director of UA at Zynga
• Yarin Ben Yaakov, Product Manager at Candivore (creators of Match Masters)
• David Reyneke, Founder of Wallabout Collective

Here's what stood out.

The first session is a deal you're making, not a tour you're giving

The report finding that 75.8% of teams say push has the greatest impact on retention in the first 30 days after install tends to get filed away as validation and moved on from. The panel's point was that it's actually the starting premise for a harder question: what does that window actually require of you?

Yarin, whose team recently overhauled Match Masters' onboarding experience, made the distinction between educating users and qualifying them, and why the difference matters more now than it used to.

"We understood something: the players coming in are different types. Some are experienced in casual games. Others are completely new to it. Because of that, we wanted to give the player the opportunity to either play the tutorial or ignore it completely. And then we want to address that when we talk to them through push. If a player ignored the tutorial, it doesn't make sense to mention it in the message. But if they started it or completed it, it makes a lot of sense to talk about the mechanics they learned… rather than sending a generic push notification that doesn't relate to what they experienced."

Baris framed his team's approach similarly, except the sequencing starts even earlier, at the ad level.

"We have ads that cater toward different motivations. Some people like the challenge. We try to create an experience that's a bit more challenge-focused in the beginning of their journey. And then we try to shift that into the communication as well. We try to showcase areas where users can actually gain value from push notifications before we ask them to opt in. In this way, we can create an initially transactional relationship, where we're sending notifications that are beneficial to the user."

David's version of this came from client work, specifically a meal plan app called Homemade Method that solved its content problem by not trying to create content at all.

"Most of their content is inspired by what's being generated by their own Facebook community. Instead of trying to figure out what content to build, they're basically using the community to build the content strategy for them. You insert yourself into a conversation that's already going on."

The through-line:

The most effective early experiences are the ones that reflect the user back to themselves. Whether it's matching your first push to what someone actually did in their first session, building your ad creative around the motivation that brought them in, or letting your own community tell you what content to make, the teams seeing the strongest early retention are mirroring rather than broadcasting.

Segmentation is both an inclusion and an exclusion problem

Our data shows behavior-triggered messages outperform standard sends by 4–9x on CTR, but that stat tends to focus attention on the messages you should be sending. What the panel spent more time on was the messages you should be holding back.

Baris described his team's approach to segmentation in terms that flipped the default framing:

"One great approach to segmentation is using dynamic tags that are constantly being updated based on player activity and preferences. That way we can identify what the player engages with the most and create segments around that.

But on the other hand, we can identify what they least engage with, and we would know that a push notification about something they don't interact with won't deliver value. So we can exclude them from receiving that push. It's not only about including players. It's also about excluding players we know it's not relevant for."

Yarin added a nuance that doesn't come up often enough: timing against global live events.

"Our game operates globally. If we start an event for the player base, we start it for all of them at the same time. The push notification tells a player there's an event going on. But the event might start when it's the middle of the night for that player. So we have to be cautious and suppress those notifications for players in that situation.

Even though it's relevant in the moment, it might not be relevant for them, because they're sleeping. And if we send it later, they might have already woken up and seen the event anyway."

David summed up the longer-term version of the same problem: what happens when your segment definitions stop reflecting reality.

"Over time, there are people who have been around for ages and there are people that have been around just for a year, and their initial experience with the app is drastically different. Some people love old school programs. Others are more interested in trying new things. We're constantly trying to re-evaluate the audience and figure out new ways to offer different things that might be more personalized, and just being realistic that not everyone is going to be the same 'active customer.'"

The through-line:

A segment is only as good as its exit logic. Knowing who to message matters, but knowing who to leave alone (and when) is what separates relevant communication from the kind that quietly trains users to tune you out.

AI is an amplifier. Don't ask it to be the voice.

Our report found that 65% of teams are experimenting with AI in their messaging workflows, but only 8% have fully operationalized it. The panel's experiences help explain the gap.

David, whose agency has shifted from a marketing shop to what he now calls a "general tech solutions consultancy," described what AI has genuinely changed, and what it hasn't.

"We are able to cast a much wider net and test a lot more variants and figure out where our human-limited amount of time should actually be spent adding the most value. But I don't think AI is as useful for building full copy and creative. It should be used more as a signal, something that supercharges your efforts. It can't be fully depended on as a marketing function. We still need a human touch to everything, or it comes off as very disingenuous when customers start to feel too much AI coming at them."

Yarin's team uses it specifically to solve a problem that's more practical than philosophical: notification blindness.

"If we use the same variety of content, the player might develop some sort of blindness to it — they see the same message type or the same text over and over, and at some point they'll ignore it. One way to mitigate that is to create a larger variety. It's either you hire a full-time copywriter and a graphic artist, or you use AI tools to enhance.

But make sure there's a human somewhere along the funnel to make sure it stays relevant and doesn't lose context. It can also be quite risky in localizing. If you don't know what you're reading, you don't know how it relates to what you're trying to say."

Baris described the same logic applied to localization (using AI to move fast on testing without waiting two weeks for a translation agency) while flagging the same caveat: human review is non-negotiable when you can't verify the output yourself.

On LLMs specifically, David made the case for using them as a data activation layer rather than a content layer.

"We're using LLMs to query the data we're already collecting on our customers to create much more hyper-targeted smaller segments, and pretty much anyone on the team is empowered to do that now. There's not as much back-and-forth with an engineer or data scientist. Everyone is able to work across all these different functions and apply it to their specialty."

The through-line

The teams getting the most out of AI are using it to make better decisions about what to test, who to target, and where human judgment is actually needed.

The bet everyone is making on the next 12 months

We closed by asking each panelist the same question: if you had to bet on one thing that will matter most for engagement over the next year, what would it be?

They all landed in the same place, each from a slightly different angle.

David:

"I think we finally have tools to make personalization a reality and to scale it. The way that you treat each individual customer is going to be how you take advantage of the future of your business."

Baris:

"Technology changes quite fast, but human nature and how we behave doesn't change as fast. At the end of the day, what we try to do is build a relationship with the user. The relationship is about first providing value and earning trust, and then becoming a place they choose to spend their time, because everybody has limited time. It's a privilege to have people come to your game and spend an afternoon there. Personalization is how you show them you understand the value of that."

Yarin, who spent much of the session talking about the mechanics of onboarding and segmentation, pointed to something more fundamental when asked to zoom out:

"The one thing that is most relevant to all players is that they are primed to look for gifts, freebies, discounts… whatever free thing they can get. This is the one thing that always delivers on the promise. Use it as much as possible."

It's a reminder that even as the infrastructure around messaging gets more sophisticated, the reason users engage hasn't changed much. They want something that's worth their time.

Want to go deeper on the data behind these conversations? Download the 2026 State of Customer Engagement Report for the full benchmarks, trend analysis, and channel-by-channel breakdowns. Or watch the full webinar to hear the panel discussion in their own words.