In 2026, I expect we’ll see big shifts in a couple key areas: how the industry is thinking about AI as a “right now” technology, and how merchants are choosing where to go for payments.
On the AI front, I think reality is starting to set in. AI has dominated the payments conversation for the past few years, showing up in nearly every roadmap and press release. The large language models (LLMs) powering tools like ChatGPT have been soaking up billions in venture capital, with the promise of world-changing breakthroughs always just over the horizon.
But I think 2026 will be the year we take a collective breather — when we start to realize what today’s AI can (and can’t) do, and wait for the next major leap beyond LLMs before the hype ramps up again.
The second shift is on the payments side. The latest generation of embeddable Software-as-a-Service (SaaS) payment products is becoming so capable and user-friendly that they’re starting to rival (or even outshine) the big-name platforms like Stripe and PayPal.
Today, more than 50% of merchants already get their payment processing services directly through a software platform. Soon, we’ll see operational software take a greater lead as the first stop for merchants who want to add payments without the complexity. In other words, truly turnkey SaaS payments are about to hit their stride.
2026 Will See the AI Frenzy Fade
The Industry Will Settle Into a More Balanced and Realistic View of AI Capabilities
I won’t go as far as to say the AI bubble is going to burst in 2026, but I do think we’re about to see a serious pullback on all the AI hype. The payments industry is no exception.
We’ve seen widespread deployment of AI tools in payments, in both internal and customer-facing applications. In some instances, like fraud prevention, it’s been great and offers a lot of value to merchants. But in many cases, the technology and tools just aren’t ready for the real world. That, or they’re solving problems people don’t really need solved.
Before I go any further, I want to make it clear that AI is incredible, and what it can do is already very impressive. It will be a highly transformative technology in the long run and will likely impact most, if not all, areas of our lives.
But it’s been three years since the initial public release of ChatGPT. I think we’re now at a point where we can clearly see the limitations of generative AI and the large language models that currently dominate the space, as well as the diminishing returns on new updates despite the tens of billions of dollars being poured into development each year.
The Nature of LLMs Limits Their Effectiveness in Engineering and Payments Contexts
Anyone who has used an LLM like ChatGPT, Gemini or Claude knows how often they hallucinate. Ask the same question ten times and you’ll get ten different answers, and two of them will be wrong. If the work one of your employees produced was wrong ten to twenty percent of the time, you would never accept that. Yet, that’s where we seem to be stuck with LLMs.
GPT-5 promised significantly reduced hallucinations, but it often accomplishes that by simply not answering. When it does answer, it continues to hallucinate way too often. And that won’t change, because this is not a bug — it’s an inherent part of what these models are. They’re prediction engines that operate on probabilities; they can’t actually think. They will always hallucinate. And that severely limits their usefulness in any task where high accuracy is a must.
And in the payments industry in particular, high accuracy is a must.
On top of that, the data problem makes things even harder. These models are trained on the internet, which, as we’re all aware, is a messy mix of truth, half-truths and outright misinformation. LLMs don’t understand which is which; they simply learn patterns. In other words, they learn statistical correlations, not truth. Teaching a model to separate fact from fiction at an internet scale is a problem we simply don’t have the technology to solve yet.
Imagine studying for an exam where a significant amount of your textbook has factual errors — even with perfect recall, your answers would still be unreliable. Similar to the challenge of operating based on probabilities, until we can curate or verify data at scale, hallucinations will remain baked into how these systems learn and respond.
Generative AI is also extremely expensive. We’ve already had to shelve some AI-based initiatives because the costs were too high to justify, and across all industries, research shows 95% of AI investments have failed to generate a return. For years, the mantra has been that costs will come down, but they haven’t. And with the big AI companies bleeding staggering amounts of money each year, the reality is that once the VC subsidization is turned off, the cost will probably go up.
We’re About to See the Industry Downgrade Near-Term Expectations for AI
The problem isn’t that progress in AI isn’t impressive – it really is. It’s that we’ve created a global narrative about how quickly it’s going to disrupt everything — and that just isn’t true. There’s plenty of evidence to suggest that AI is a normal technology, like any other, and it will evolve, grow and find use cases through a normal cycle that takes many years.
The dot-com bubble didn’t happen because everyone was wrong about the impact the internet would have. It happened because they were wrong about the timing. The speed of AI adoption is often pointed to as evidence that “this time is different,” but the rapid deceleration in returns on each new model shows us that it probably isn’t.
AI will be widely impactful, but we may not see the true payoff for another ten years because it’s going to require transformative shifts from where we are today. Because of that, I think 2026 will be the year we start to see expectations reset — and hopefully a shift back to days of having to prove something works before selling people on hype.
Instead of chasing quick wins or “any day now” breakthroughs, timelines will stretch out to five or even ten years. Companies won’t stop experimenting with or building on AI platforms, but they’ll start using them more intentionally, treating AI as a tool that makes sense in the right context, not something to force into every problem. In my view, that’s a healthy correction — and one that’ll ultimately benefit the entire industry.
Humanless Onboarding: SaaS Companies Will Start Onboarding Merchants Without Humans
AI hasn’t replaced humans at scale the way many predicted, and it’s not yet making our day-to-day lives as effortless as we hoped. But one area where I am seeing real progress, in both simplicity and efficiency, is SaaS-driven payments. Which brings me to my second prediction: the rise of software-led embedded payments as the new, low-friction starting point for businesses.
Most merchants today rely on some form of operational software. While there’s no question that SaaS-driven payments represent the future of the industry, right now, there’s still a noticeable gap between how easy it is for a business to start processing payments through their SaaS platform versus the big, household-name providers. We’re about to see that gap close dramatically.
Soon, I believe we’re going to see 50% or more of all SaaS-based merchants onboarding to payment services with zero humans in the loop. They’ll sign up through their operational software platform, and it’ll just work right away. No fuss, no technical difficulties; just opt in and go.
The SaaS-focused payment products that we’re currently developing are getting so good, so easy to integrate and so frictionless that the end merchants won’t need that helping hand anymore. The sign-up, the underwriting, the deployment of the gateway and the value-added tools — they’ll all be so seamlessly integrated into the software that, for the end business user, payments will be as simple as sign-up, toggle it on, and start getting paid.
That has huge implications for the industry because, if SaaS companies can offer the same ease and simplicity as the huge platforms and the lower pricing of a more traditional merchant account, why would any merchant say no? I think we’re on the verge of a big shift where SaaS-based payment services become the go-to “easy” choice for most merchants, instead of the big-name platforms.
To learn more about how we’re enabling our partners to stay at the forefront of payments innovation, reach out to a member of our team.