A growing market narrative suggests that AI will “eat” specialist software. In wealth management, finance, marketing and beyond, the fear is that general-purpose AI agents will displace SaaS platforms, capturing the highest-value user interactions and leaving software providers hollowed out.

It’s a compelling story but in payments-enabled software, it’s almost certainly wrong.

In fact, in the emerging world of AI-driven and agentic shopping, SaaS platforms that have invested heavily in embedded finance are not the losers, they are set to become among the best-positioned players in the ecosystem.

To understand why, it’s worth stepping back and asking a simple question: what does agentic commerce actually require to work at scale?

The three things AI shopping can’t do without – and the fourth that changes the economics

For AI agents to research, recommend and complete purchases autonomously, four things are non-negotiable.

First, reliable and structured SKU-level data.

Platforms like Shopify, Lightspeed and Toast already sit on vast, structured product and menu datasets. This isn’t scraped, or inferred information – it’s canonical data that merchants already actively maintain to run their day to day online and in-store businesses. It’s often real time updated, and immediately usable by AI agents and can be enriched over time. General AI platforms don’t control this data; embedded SaaS platforms do. Those who own and control data are king makers in the world of AI.

Second, trusted and monitored sellers.

Agentic commerce only works if buyers, human or AI, can trust the counterparty. Embedded SaaS platforms have already done the hard work: onboarding merchants, conducting due diligence, understanding trading history, monitoring fraud and managing risk on an ongoing basis. This trust infrastructure is invisible, expensive to build, and extremely difficult for a standalone AI platform to replicate. If consumers are going to use these new AI platform, they need to trust the shopping recommendations they receive, and SaaS platforms can hold the key to millions of trusted sellers. 

In a world where consumers are using agents not just as standalone recommendation engines (e.g. Claude or ChatGPT), but also embedded in large platforms (e.g. Amazon’s Rufus) the issue of trust becomes even more important. Large brands like Booking.com or Amazon are looking at agentic commerce as a way for consumers to buy outside their supply chains. Indeed Amazon already has a product called “Buy for me” which does exactly this. These platforms with trusted consumer brands are not going to want to scrape content from the internet from sellers with limited credentials. They are going to want to use trusted third partry networks. These come ready made within SaaS platforms.

Third, deep integration with modern payment infrastructure.

Most leading vertical SaaS platforms are already plugged into payment providers like Stripe and Adyen – firms that are at the forefront of tokenisation, real-time payments and emerging agentic standards. A wide range of payment types, fraud monitoring of consumers and their agents, dispute handling, refunds and data security are not “nice to haves” in AI commerce; they are table stakes.

Some platforms, such as Shopify, have also invested in their own consumer payment brands (Shop Pay) which will help build trust and enable faster in-agent checkout. Indeed Google’s UCP standard currently only functions with stored payment credentials.

Fourth, the ability to aggregate millions of sellers and route them to many agents.

The final requirement and the one that fundamentally reshapes platform economics, is the ability to bring together millions of smaller sellers who want to use AI agents to reach new customers and connect them to a growing ecosystem of agents and surfaces.

We’re already seeing early signs of this shift. Shopify, for example, is evolving from “I help you sell online” to “I bring you customers and help you convert them.” Payments might generate 3-4% in fees; demand generation and conversion can double that economics. Lightspeed and others have similar opportunities ahead of them.

This is a powerful positive feedback loop: embedded finance enables agentic commerce, and agentic commerce increases the strategic importance and monetisation potential of software platforms.

In an agent-driven world, distribution becomes as valuable as infrastructure. Embedded SaaS platforms are uniquely positioned to act as neutral marketplaces between sellers and agents – not by owning the customer outright, but by enabling trusted, scalable access to demand wherever AI shopping happens.

Embedded finance becomes the distribution platform

The market’s current pessimism also misses a crucial second-order effect: AI accelerates the value of embedded finance.

Software platforms that successfully embedded payments did more than add a revenue line. They embedded themselves into the commercial flow of their customers’ businesses. That same positioning now allows them to become the natural content and delivery layer for AI-driven demand.

If not SaaS, then who is at risk?

If AI doesn’t ‘eat’ embedded SaaS, it’s fair to ask: who does it disrupt?

The uncomfortable answer may be marketplaces which currently generate c.50% of eCommerce volume globally*. 

In a world where an AI agent can discover, filter and transact directly with trusted merchants using platforms like Shopify or Toast, the value of intermediaries whose primary role is aggregation comes under pressure. Do you need a food delivery marketplace if your AI agent can order directly from local restaurants, using their existing digital menus, payment rails and fulfilment options provided by players like Toast?

This doesn’t mean marketplaces disappear overnight. Logistics, consumer trust and scale still matter greatly – Amazon is the obvious example. But in categories like local food delivery, where proximity matters and supply is already digitised, the long-term balance of power is far from settled.

It’s no coincidence that marketplaces are racing to embed their own AI agents. They see the same risk.

A market rout looking for a rethink

Today’s market reaction – selling off SaaS on the assumption that AI makes them obsolete – feels misplaced. In many cases, the opposite is true. The threat of AI may be real in many B2B SaaS use cases where they have not built the exclusive, content rich relationships we have been discussing here, but perhaps it’s something they should consider building.

*Source: Future of Marketplaces report by Edge by Ascential