Atlassian and HubSpot are joining the growing list of software vendors abandoning flat-fee pricing for AI features, according to The Information. The shift signals that the early “AI included for free” era is winding down, as vendors realize that inference costs and customer value vary too much for a single flat rate to work.
The Information reports that both companies are moving toward usage-based or tiered models, following the path already taken by Salesforce, ServiceNow, and Microsoft over the past year.
What’s actually changing
For most of 2024 and early 2025, SaaS companies bundled AI features into existing subscriptions or sold them as a flat add-on. The logic was simple: get AI in front of customers fast, drive adoption, sort out pricing later. That worked when usage was light and inference costs were manageable.
It doesn’t work anymore.
Heavy AI users now generate compute costs that blow past what a flat seat fee can cover. Light users feel like they’re subsidizing power users they don’t compete with. Vendors are caught in the middle, eating margin on one end and losing perceived value on the other.
Why this matters now
AI feature adoption has scaled fast enough that the economics finally bite. A flat $20 per seat AI add-on looks fine when 10% of users actually touch the tool. When 60% of seats start running daily AI workflows, the math collapses.
There’s also a competitive angle. Vendors who price by outcome or consumption can charge proportionally to value delivered. That means they capture upside from heavy users without scaring off small teams. Better long-term business than flat-fee bundles.
The pricing models taking over
Three patterns are showing up across the industry:
- Consumption-based: pay per AI action, message, or compute unit. Microsoft Copilot Studio and Salesforce Agentforce both work this way.
- Tiered seats: light AI usage in standard seats, heavy AI workflows gated behind higher tiers.
- Outcome-based: pay only when the AI completes a defined task. ServiceNow has tested this with agent workflows.
Atlassian and HubSpot are reportedly mixing these approaches rather than picking one. That hybrid model is probably where most of the industry lands by end of 2026.
What this means for buyers
If you’re procuring AI-enabled SaaS in 2026, the pricing page is going to look different than it did six months ago. A few practical moves:
- Audit your actual AI usage before renewal. Light users may save money on consumption pricing. Heavy users will pay more under almost any new model.
- Push for usage caps or volume commitments. Most vendors will negotiate ceilings if you commit to a floor.
- Watch for hidden inflation. “Free AI” is being replaced with paid AI, and the new bundle price often includes credits that run out fast.
- Model the worst case. If your team adopts AI features faster than expected, what does the bill actually look like?
For SaaS founders and product leaders, the takeaway is sharper. Flat-fee AI is a pricing trap. It looks customer-friendly on the way in and crushes margins on the way out. Vendors who structured pricing around consumption from day one are going to look a lot smarter in their next earnings call.
The forecast
Expect every major SaaS vendor with AI features to announce some form of consumption or outcome-based pricing by mid-2027. The flat-fee experiment was a land grab, and the land is grabbed. Now comes the part where vendors actually need to make money on this stuff.
Atlassian and HubSpot moving now suggests the shift is already past the tipping point. Companies still on flat fees aren’t holding the line out of strategy. They’re just slower to react.
More details on the specific pricing changes are available at The Information.