AI Agents are not going to kill your SaaS
A prediction on where the SaaS and AI landscape is headed.
LLMs are stochastic parrots. They just predict the next most likely word. And the most likely token to follow “AI” these days is, without a doubt, “Agents”.
AI agents are hailed as the next quantum leap in the evolution towards building SkyNet AGI.
An agent is basically a program that can interpret input with an LLM and has access to tools to actually do something with that input. Instead of ChatGPT drafting you an email, an agent could send emails on your behalf. You could ask your agent to “send friendly reminders to fill out that Doodle”, and it could read your latest Doodle, see who didn’t reply yet and then ask the Gmail agent to send those reminders to the right people. Rather than text-in-text-out, an agent converts input into actions.
It’s one of those big promises of AI. Agents should allow us to delegate most manual office work to machines, all in natural language.
Satya Nadella had a few nuanced insights into how AI Agents would reshape software architecture, which the internet famously summarized as “SaaS is dead”. It’s a bad take, but it’s easy to see how people can come to that conclusion.
SaaS products are not always user-friendly. They have built-in limitations designed to drive more traffic or to upsell functionality. Take Substack, for example. There is no way to create a campaign that would automatically create a daily Note to promote one of your articles. Marketeers and growth-hackers would love such a feature, but Substack doesn’t offer it. They don’t want you to use the product like that.
SaaS products are in control of how you access their data. They support the behaviour they want to see, while making unwanted behaviour hard. Skyscanner doesn’t allow you to search for next week’s cheapest flight out of Brussels. But you could, in theory, try every possible combination, store the results in a spreadsheet and sort them by price. The rationale behind “SaaS is dead” is that agents will do those kinds of mind-numbing tasks to work around the limitations.
But there’s an obvious issue with that line of thinking. Playing around with agents is already possible today with tools like Manus. You can ask Manus to look for the cheapest flights, and it will go through Skyscanner and its many competitors, scouring for that deal. But if you do that, you’ll notice that it doesn’t work. Or better, that it doesn’t work anymore.
“Find the cheapest flight” is the Hello World of AI Agents. Everyone is trying that. And as a result, these SaaS products have set up their defences. Elaborate CAPTCHAs, rate limiting, IP blocking and even AI Labyrinths to confuse the bots. While Manus is a fantastic implementation, it’s all but useless for finding the cheapest flight.
There’s already an arms race going on where those impacted by agentic swarms are finding ways to tackle them. Since the SaaS products ultimately control which data gets exposed, they will win the race.
Does that mean AI agents are a dud? Not at all! Companies will offer agents to their customers who can use their tools in line with the product vision. And that’s the crucial difference. These agents will not be rogue actors under the control of script kiddies. They will be SaaS products, with subscription models and the ability to upsell.
An agent, as a combination of product feature and customer touchpoint, will be a golden opportunity for SaaS companies.
If you’re building a SaaS, which services could an AI agent offer your users?