When Your Subscription Becomes a Meter: Rethinking Build vs. Buy as SaaS Moves to Usage-Based Pricing
Evaluating software used to be simple: count the users, multiply by the monthly price per seat, and you had a budget. That math is breaking down. Atlassian has announced that from December 3, 2026, it will start metering AI features and automation steps in Jira and Confluence, and more and more customer service tools now charge each time AI successfully resolves a case. SaaS usage-based pricing turns your subscription into a meter, and it means the old question of whether to build or keep renting needs a fresh calculation.
SaaS usage-based pricing arrives: Atlassian's two meters
According to a summary by Atlassian partner ReleaseTEAM, with figures as of September 24, 2026, two kinds of usage start being billed on December 3:
- Rovo credits. Credits are consumed based on the type and complexity of each AI interaction; some search, definition and summary functions may not consume credits. For Jira and Confluence, Standard plans include 25 credits per user per month, Premium 70 and Enterprise 150, pooled across the organization. Overage costs $0.01 per credit.
- Automation steps. Usage used to be counted by rule runs; now every executed step counts, including triggers, conditions, actions, branches and loops. Overage is $0.50 per 1,000 steps.
Two details deserve attention. Extra usage is switched on by default, though admins can set caps or turn it off. But if allowances run out and extra usage is off, some automated processes may stop until capacity is available again.
It's not just one vendor: software pricing is moving from seats to usage
Atlassian is not an outlier. In a May 2026 analysis, Futurum Group found that buyer preferences are shifting: 43% of enterprise buyers prefer consumption-based pricing, 27% prefer outcome-based models, and fewer than one in five still prefer classic per-user pricing.
Vendors are following. Futurum notes that Zendesk now charges for successful AI-driven issue resolutions rather than AI seat licenses, Intercom uses a similar model for its Fin AI agent, and Salesforce tracks AI actions in Agentforce through consumption credits.
The reason is easy to see. AI lets one person oversee a great deal of automated work, so "how many people use it" no longer reflects the value software creates, or the compute costs vendors carry.
Bills get harder to predict: three risks to watch
Usage-based pricing isn't necessarily more expensive, but it is always harder to forecast.
1. The better it works, the more you pay
The better your automations are designed and the more your team relies on AI, the higher your usage. A tool meant to cut labor costs can turn into a bill that climbs every month.
2. One workflow may be metered twice
A single workflow can draw on both automation steps and AI credits at the same time. Unless someone maps the cost of each step, it is hard to know where the money goes.
3. What stops when the allowance runs out?
If you disable overages to control spend, you need to know which workflows would halt and whether that affects customers or operations. That is not just a finance issue; it is an operational risk.
Recalculating build vs. buy: four questions
We covered the basic framework in Custom Software vs SaaS in 2026. As usage-based pricing spreads, add these four questions:
- Will your usage grow sharply with the business? If a workflow scales with orders or customers, long-term usage fees may exceed a one-time development cost.
- Is this workflow part of your competitive edge? Generic functions are usually cheapest to rent. But for the processes that set you apart, building puts both cost and experience in your hands.
- Do you only need a small slice? Many companies use one feature of a large platform yet pay under the whole system's pricing rules. A lightweight custom tool for that slice is often simpler.
- Building doesn't make AI free. Custom software still pays model usage fees every month. See The Line Missing From Your Quote to estimate post-launch costs before you commit.
The best answer is usually hybrid: keep renting general collaboration, accounting and help desk platforms, and turn the workflows that are tightly tied to your business, and that scale with it, into custom systems whose costs you control, connected through APIs.
Conclusion: do the math before you decide to build
SaaS usage-based pricing isn't a bad thing. Light users pay less, and prices track value more closely. But it hands responsibility for cost control back to the business: someone needs to understand who uses what, how much, and which workflows cost the most. Before December 3, take stock of your team's real automation and AI usage, then decide what to keep renting and what is worth building. TitApeX helps companies assess workflows, then plan and build custom systems, letting technology handle precision while your team focuses on what moves people. To rerun your software math, explore our services or talk to us directly.
Comments