Is usage-based pricing becoming the norm for AI tools?
Hey everyone,
I've built my product around traditional SaaS pricing (monthly tiers), but I’m starting to wonder if that model is getting outdated, especially with more AI-powered and compute-heavy tools entering the market.
That shift requires real architectural changes, instrumentation, metering, billing logic, and UI changes, not just pricing tweaks. It’s something I’m starting to seriously think about for my own product.
In particular, AI usage has real COGs (every prompt costs money), and I’m seeing more platforms experimenting with usage-based models, or hybrids like “SaaS base + usage + overage.”
For those of you building AI or compute-intensive tools:
Are you sticking with SaaS pricing?
Have you considered switching to usage-based or hybrid models?
Is it helping or hurting conversions?
Would love to hear what others are doing and whether you're seeing buyer preferences shift, too.
Replies
Honestly the usage costs are brutal right now. I keep telling myself its an investment that'll pay back in productivity or whatever but some months I look at the bill and just... yeah. The tricky part is once you build workflows around these tools its hard to go back so you just keep paying and hoping the value compounds faster than the invoices
@gunhorizon what type of features in your products have you noticed the largest usage costs?
@jake_friedberg Right now its Clawdbot automation eating me alive (in the name of productivity) but I am doing an app that generates your selfie in anime / manga style and that ain't cheap either
I think it might, but that only works for businesses and individuals who are willing to do this in my opinion, so people who either are tech savy or see so much value that inconsistent costs are okay for them since the ROI is much larger.
@thomasp0 yeah, its a big change so if a company was going to change their billing, it needs to be well thought through.
I’m personally torn. Usage-based is attractive because it feels aligned with how resources are consumed. But I agree that it complicates the billing and can make forecasting tough. I’ve seen some hybrid models where users get some “free” usage and pay more as they scale — that seems to be a good balance.
@prudens_moulton What is your product, is AI centric to its value or are you implementing it as an add-on?
Tonkotsu
Very interesting and challenging part of GTM right now. I agree that standard subscription pricing has lots of benefits: understandable, familiar, predictable - all very good things. But with inference costs being relatively high, you might need to set the pricing higher than folks expect with traditional saas.
With usage pricing, another challenge (this is most obvious in the coding space) is if the frontier labs offer subsidized consumer pricing (as Anthropic is doing). You can't compete with them on this, and it makes the consumer -> enterprise upsell path challenging.
@derekattonkotsu yes, are you going through this process right now with your project?
Tonkotsu
@jake_friedberg yeah, definitely thinking about it. Enterprise is willing to pay full inference pricing at scale because the economics work out. Hobbyists aren't (no one's going to pay $500 a month for their fun side project) and they're the front door before upselling into enterprise.
It really depends on the tool.
In my case, I spent weeks figuring out the best pricing to cover costs and provide value.
I ended up with a "baseline" offering plus "pay-as-you-go" for extra usage.
Personally, I dislike credit systems; I prefer to know exactly how much I'm spending and what I'm getting. I think others feel the same way.
It’s fascinating! I recently launched bracework, and it employs the exact approach I used … a hybrid monthly fee plus usage-based pricing.
Token cost money, and this model prevents super users from abusing the system. It educates them about the actual cost of using an assistant. In my case, Bracework is an AI documentation assistant that helps field service professionals get estimates, invoices, scope of work, or any other documents they need to win a job faster. It can do this in less than 45 seconds by texting a photo, voice, or text note from their phone.
@olivier_madel1 nice Olivier! Just took a look at your site, it looks very well thought out and put together, did you get feedback on the pricing before implementing it, or did you just go for it?
@jake_friedberg @jake_friedberg, I realized how much it costs me to use AI or call an LLM for my service after doing some calculations with Claude code. We came to the conclusion that this was the best avenue to pursue for profitability. Instead of saying, “You’ve used your credit for the month,” I could have gone with a SaaS pricing model. Now, I could have gone with pricing per token, which would make sense. I priced fairly one usage cost to generate several documents. So, the fee is per job. It’s still manageable per unit cost. And in the language of my ICP, this makes sense. Jobs are something they understand, not tokens or whatever. The math is easy. I spend X to generate several documents to win or close a job. The value proposition is that the user wins all the time while having a great experience with the LLM providing the service: a document.
I think the same!
my SaaS Ptio has usage based pricing based on token usage plus a small margin to cover other transactional costs
We went the complete opposite direction with Geode (Monthly / Yearly / Lifetime Deal).
Because our AI runs strictly on-device (using the user's local hardware/Neural Engine), we don't have the massive per-token cloud costs that force other AI tools into usage-based models. Our marginal cost per user is near zero regardless of how much they use it.
We found that avoiding "meter anxiety" is a huge selling point. Users prefer knowing they can transcribe 5 hours or 50 hours of meetings for the same flat price (or a one-time fee), rather than worrying about running out of credits.
For us, Local Architecture = Flat/Lifetime Pricing. It’s a major differentiator against the cloud wrappers.
@luciawantstoknowall the LTDs are interesting and something I've read, which has varying success. But it sounds like you've implemented this wisely, how is the LTD working out for you?
Dirac
I reckon the reason a lot of SaaS apps use usage-based is because with ai, the spend can fluxuate very rapidly. So one month the app would lose money from the subscription, and the other month the user wouldn't use the full value of the app. So the usage-based is a perfect balance.
For me, usage based sees a lot more trying-it-out users, versus standard SaaS. So theres more conversions.
@peterz_shu Yes solid point and as was said in this thread a few times its really the balance of both. Pricing can get messy I personally am a fan of the freemium model + upgraded for MRC with add-ons of usage. But its definitely a mouthful depending on your business strategy.
Dokably
I see that the usage-model system has been there for some time already. I build a productivity tool Dokably and in this domain (Notion, Trello, Clickup) there is a mixture of subscription base model (pay and get features) and usage-based.
@sasha_dikan I'm familiar with all of those tools, but am not familiar about their hybrid pricing models. How are these tools using AI to justify a new added usage cost?