MPE Berlin Recap: The gap in payments is becoming harder to ignore

I recently spent three days at MPE Berlin 2026 across a mix of panels, keynotes, conversations and many meetings with merchants, payment providers and partners.

What came through consistently was not a lack of technology or ideas, but a shared problem: businesses that accept payments online are still making critical decisions without certainty. Fraud continues to rise, but at the same time, legitimate customers are being declined, and revenue is being lost because risk systems can predict behaviour, but they cannot prove intent.

It was clear to me that the industry is aligned on the problem, but not yet on the solution.

Prediction has a limit

A lot of the conversation was focused on AI, data and behavioural analysis. That makes sense. These tools are getting more sophisticated and they are helping businesses make better decisions about risk. But they are still, fundamentally, making predictions. And prediction has limitations.

There will always be a point where the system is not sure, and so the safest option is to decline the payment. This is where significant cost sits. Not just in fraud, but in legitimate customers being turned away by false positives. This isn’t just a risk problem - it’s a revenue problem measured in billions of declined legitimate transactions.

The trade-offs the industry is still living with

Across multiple sessions, including those focused on fraud prevention and cyber resilience, the same core issues came up repeatedly:

• false positives

• increasing fraud sophistication

• lost sales

• eroding customer trust

These are the trade-offs that online merchants are still living with.

In trying to reduce fraud, businesses are creating friction and turning away genuine customers. In trying to improve acceptance, they expose themselves to more risk. And at the heart of it all, most systems are trying to answer the same question: how likely is this transaction to be genuine?

But ‘likely’ is not the same as knowing. And when you don’t know, you either accept risk or you decline the transaction. Both have a material cost.

Agentic commerce will amplify the issue

One of the dominant topics across the conference was agentic commerce. There’s a lot of energy behind it, and understandably so.

My view is that we are probably overestimating the short-term impact but underestimating what it becomes over time. It will take a while before it is truly mainstream, but directionally, it matters. What was interesting is that even in those future focused discussions, the same issue kept surfacing - fraud.

As transactions begin to be initiated by agents rather than individuals, the question of trust becomes even more complex. The industry is already dealing with significant fraud today, and that risk is only expected to grow.

Much of the thinking comes back to better prediction, which looks like more data and intelligence. These will help, but they don’t remove the underlying problem because once again, predicting isn’t the same as knowing.

Moving from prediction to certainty

The solution to all this isn’t in awareness, effort, or indeed more technology. It’s in how we move from estimating risk to establishing certainty. There will always be a place for AI and behavioural analysis. They are critical tools, but they aren’t designed to provide definitive answers in moments of uncertainty.

Those moments need a different approach.

Instead of asking the system to make a better guess, the question becomes how to confirm, with confidence, that a customer is genuine. This is not a new idea. In physical commerce, we already solve this problem with card present transactions where we authenticate the customer at the point of payment.

What feels clear is that digital commerce is starting to move in that same direction. Burbank’s CPoI® (Card Present over Internet®) is an example of how authentication can be introduced at the point where risk systems are uncertain, rather than relying solely on increasingly complex prediction.

Until that shift happens, the trade-offs I mentioned above will remain.

What comes next?

MPE made one thing very clear: the problem is well known. But as payments continue to evolve, especially with the introduction of new models like agentic commerce, the cost of not solving it properly will only become more visible.

The next phase isn’t just getting better at spotting risk; it’s about resolving it with certainty, in a way that reflects how we already trust payments in the physical world.