A friend of mine runs a small SaaS company project management software, maybe a few thousand paying customers, nothing flashy. Last year he told me something that stuck with me: “I used to sell seats. Now I’m not totally sure what I’m selling.” He wasn’t being dramatic. He’d just noticed that half his customers weren’t logging in as often anymore, because an AI feature inside the product was doing a chunk of the work they used to do by hand.
That’s the quiet shift happening across the SaaS business model right now, and it’s a bigger deal than most of the AI-feature-announcement posts make it sound.
It Was a Straightforward SaaS Arrangement that Was Easy
The basic proposition for years has been that for a monthly fee, pay up for a seat and you’ll get software that will help you with a particular task. The more people that are on the team, the more seats, the more money. It was successful because it was something that someone actively used, logged in, clicked around, and did the work themselves, but with better tools than a spreadsheet.
That model assumed humans were doing most of the actual work, with software just making it faster. AI breaks that assumption in a way a lot of SaaS companies are still figuring out how to handle.
When the software starts doing the job itself
My friend’s product added an AI feature that automatically organizes tasks, flags overdue work, and drafts status updates without someone building them manually. Customers loved it. Usage of the manual features dropped. And that’s where it got complicated: his pricing was built around seats and logins, not around how much actual work the software was doing on someone’s behalf.
He wasn’t alone in running into this. A lot of SaaS products are quietly facing the same math problem: if an AI feature does the work that used to require five people logging in daily, charging per seat doesn’t make sense anymore. Fewer humans need to touch the product, even though the product is doing more than it ever did.
Pricing is where this gets messy
The obvious fix sounds simple — charge based on usage or outcomes instead of seats. In practice, it’s a mess to figure out. Usage-based pricing on AI features means tracking API calls, processing time, or tasks completed, and none of that maps cleanly onto the flat monthly invoice customers are used to seeing.
My friend ended up testing a hybrid model — a smaller base seat price plus a usage charge tied to how much the AI feature actually did. Some customers liked it because light users paid less. Others hated the unpredictability of a bill that could change month to month. One customer emailed asking for a spending cap because they’d been burned by a different tool’s surprise usage bill a year earlier a totally fair ask, but not something his billing system was ever built to handle. There’s no clean answer yet, and talking to a few other founders in similar spots, nobody seems to have fully cracked it either.
Customer support got weirder, not easier
Here’s something that doesn’t come up enough: AI features inside a SaaS product changed what support tickets look like. It’s not “how do I use this button” anymore. It’s “why did the AI do this instead of that,” which is a much harder question to answer, because sometimes the honest answer is that the model made a judgment call nobody can fully explain.
My friend’s support team had to learn an entirely different way of troubleshooting — less about walking someone through steps, more about explaining probabilistic behavior to someone who just wants a definite answer. That’s a real skill shift for people who were hired to answer “where’s the export button,” not “why did the system summarize this differently than last time.” He ended up pulling in someone from the product team just to sit in on support calls for a few weeks, because the old scripts genuinely didn’t cover this kind of question.
The retention story is flipping
Traditional SaaS retention logic says more logins and more feature usage mean a happier, stickier customer. AI complicates that too. A customer whose AI assistant now handles most of what they used to do manually might log in less, not because they’re unhappy, but because the product is working exactly as intended.
My friend had a minor panic when his daily active user numbers dropped after the AI rollout, assuming people were churning. Revenue stayed flat. Renewals stayed strong. People just weren’t clicking around as much, because the software had quietly started doing more of the clicking for them. He had to rebuild how his team even measures whether a customer is actually engaged.
Where this actually leaves SaaS companies
Nobody’s fully solved this yet, which is honestly a little reassuring; it means the companies figuring it out as they go aren’t behind some curve everyone else has already cleared. The common thread among the founders I’ve talked to is that pricing, support, and the whole idea of “engagement” all need rebuilding from a different starting assumption: the software itself is now doing meaningful work, not just helping a person do theirs faster.
My friend hasn’t landed on a perfect model yet. Neither has anyone else, really. But he’s stopped trying to force the AI features into the old per-seat pricing box, which at least feels like progress, even without a tidy conclusion to point to yet.