Will AI Agents Become the Internet’s Next SaaS Buyers—or Just Fuel Big Tech Bundling?

Design is Love

SubscribeSign in

Will AI Agents Become the Internet’s Next SaaS Buyers—or Just Fuel Big Tech Bundling?<br>Apparently humans with credit cards are the bottleneck.

Sal<br>Aug 18, 2026

Share

A lot of young founders today are pitching VCs this story - AI agents will be the biggest consumers of SaaS in the future. It’s an intoxicating story, because we already use agents for workflows and other parts of our daily lives. I personally use an agent to go through all my messages and inbox and sort them out into a clear To-Do list. It saves a lot of time, but in the end, I put my credit card myself.<br>If we think of the world 5 years from now, would there be cases where an agent’s tool-call essentially allows it to purchase a subscription that then enables them to do things they otherwise couldn’t have? Or would these capabilities simply be bundled into bigger software makers (the OpenAI’s the Anthropics, or other venture funded products)? Looking at the history of the software industry, bundling is definitely the way things were, and it’s a good bet to assume that history would repeat itself here.<br>This creates two scenarios that are worth exploring individually.<br>First, the world where AI agents are doing the purchasing. This assumes that the services provided by the different companies in a space are irreplicable. This is hard to believe, because if coding agents today are so helpful that Cursor is able to basically build its own version of Github, then we should assume that coding agents would be good enough to replicate most basic software fairly quickly. There are a few exceptions, and the best example I can think of is OpenClaw. If the founder of OpenClaw had made the product closed-source, it would be near impossible to copy. The IP would have made it so that companies would have spent months maybe a year (or until the product runs its lifecycle) trying to imitate the very basics like memory, skills, and other parts of it. Of course, the product here was built to be used locally, and so closed source never made sense.<br>It also assumes trust for the AI agent to make a purchase decision. This is easier to believe, because all it needs to do is tell me what the product would enable it to do, and maybe have me give the final approval.<br>The third assumption is that the internet is no longer a place to be desired, and that there are no people willing to browse it with their own eyeballs to come to a purchase decision. It’s a bold assumption, but the problem here is, a lot of the software I hear about or am curious about, is mainly because of what other people said, and not what other agents said. If I go to a housewarming in SF, meet other founders, and they give me their websites, I would most likely visit to see what their product is about and maybe try it. AI will never replace that human-to-human connection. Right now, I’m sitting in a cafe in my home country of Bahrain, but I know that physically being in SF is an incredible opportunity and an unfair advantage for anyone who has the ability to do that. The internet bridges this geographical gap for me, and assuming that the internet would no longer be a place for humans to browse might have some truth, but not completely. A new scary statistic from a talk I saw at MicroConf proves this to an extent. AI has indeed reduced redirects, and as founders we have to work harder to get our website out there. But are websites obsolete - not yet, though people increasingly just look at the AI answer and skip the whole website. Personally, I noticed AI answers are fine for a quick overview, but often times visiting someone’s website gave me more information that AI just skipped over. It’s like reading an essay. AI just doesn’t feel the same.<br>The second scenario is more imaginable. Everything gets bundled into a big company and you get the next Microsoft or Apple or whatever. Agents can do things, but because the incumbents have so much market share, it’s their agents that do most of the things on the internet. Their agents are optimized to save that extra 1% on transactions therefore making them the agent of choice. Their ecosystem—one you would have to fight to enter and operate within for your agent to perform as well as theirs. Their scale carries them forward.<br>Very easy to picture, but one part of this that I simply couldn’t predict 5 years ago is that every component - the models, the harnesses, the routers - is open source and available for anybody to build out of. The only differentiator is that the incumbents can probably get their own tokens to the market cheaper, but even startups can come up with systems to offer tokens. OpenRouter did it pretty well, so can anyone else. So if the pipes, plumbing, and water source is all up for grabs, then the only moat here is market share & money.<br>So what bet to make? I don’t think we’ll be in a world where agents are the only...