Nobody wakes up wanting your interface. They want the job done, and specialized AI is now doing it, priced by use.
By “AI model,” I mean a standalone AI offering built around a specific job, with pricing tied to usage or results. I’m not suggesting every SaaS company needs to train its own foundation model. I’m asking whether customers can buy an outcome from you without buying your entire platform.
Here’s what opened my eyes. I used to produce videos for a living. Over time, as my career evolved, I made videos less frequently and free tools like iMovie became good enough for those moments. Last month, I needed a high-quality video made in one hour. The old way would have taken a full day. So I went looking for AI.
1. Buyers want the job done, not the interface
I wasn’t looking for another monthly plan for one video, or a reason to relearn a complex editor like Premiere. The tools that came up were names I’d never heard of in the video production space: HeyGen and Hedra for AI avatars, ElevenLabs for narration, Runway for video generation. I could choose tools around the pieces of work I needed done. The pricing isn’t all pure pay-as-you-go: HeyGen and ElevenLabs, for example, combine subscriptions with credit allowances. That’s an important distinction. A usage meter inside a subscription isn’t the same as letting someone buy one result.
This isn’t only a creative-tools story. Intercom lists Fin AI Agent pricing from $0.99 per outcome. Fin can also work with help desks such as Zendesk and Salesforce. The AI capability can reach customers who haven’t bought Intercom’s help desk.
2. Quality-obsessed buyers will pay for the best
Some customers have never settled for “good enough,” because their reputation is their product. Think of creative ad agencies, interior designers, consultants, or even five-star restaurants. A result that isn’t top quality can cost them clients. But even for them, there are still things they have likely only dreamed of but had to skip out on to be more realistic.
For me, a 3D animation or a polished on-screen narrator meant hiring a specialist, waiting weeks, or blowing the budget, so those things never made it into my plans before. Specialized models change that. One video can now have a lifelike avatar, natural narration, and generated footage, each from a tool built for that one job. What I once wrote off is back in reach, and I can pay for it as I use it.
Established studios are exploring this, too. Lionsgate announced a partnership with Runway to develop a custom model using its film and television catalog. AMC Networks announced plans to use Runway for previsualization and promotional imagery. These partnerships don’t prove that every output meets the bar or that every buyer wants usage pricing. They do show companies whose reputations depend on creative quality taking specialized AI seriously.
For SaaS leaders, the lesson is simple. This buyer doesn’t want a seat license they barely use. They want the best result for the job, and they’ll pay by use to get it. Would anything on your roadmap earn that label?
3. An AI feature isn’t an AI model
It’s easy to hear “AI strategy” and add a chatbot or a few AI power-ups to the existing product. Adobe’s 2025 Creative Cloud changes, for example, paired a higher-priced Pro plan with expanded generative AI access, while also offering a Standard option. Bundling AI can be a useful strategy. But it answers a different question from the one I’m asking: can someone buy the specific job they need done without committing to the larger suite?
4. Seats are shrinking, not vanishing
A Gartner forecast cited by Deloitte puts at least 40% of enterprise SaaS spend on a path toward usage, agent, or outcome-based pricing by 2030. That’s a forecast, not a settled outcome. Bain’s analysis also points to hybrid approaches, where seat pricing and AI-related charges coexist. I wouldn’t read either as a reason to abandon subscriptions. I’d read them as a reason to test what else customers will buy.
Both models will live on. The question is whether your roadmap and your sales & marketing engine can sell both.
How it comes together
Layer one: standalone AI models for specific jobs, priced by usage, built for quality and speed. They can reach customers your flagship never could, and they work without it.
Layer two: your flagship becomes the control center, where leaders see usage, spend, and outcomes across every AI tool in a function, built on the workflows and data customers already trust.
This isn’t free money. Usage costs rise with inference—though AI operating costs should become more efficient over time—and some seat revenue may shift. That’s why I’d start with one job, not a full pivot.
Begin with customer and market research. Find the jobs where quality matters most and pay-per-use makes sense. Then ask what your customers have written off as out of reach.
Open your roadmap. Is there a single AI model on it, built for a job your most quality-obsessed customers would pay for by use? Or only features?
Image credit: Peter Stumpf / Unsplash.
