Automation Platforms Have Switched to Credits. Your Old Budget Does Not Apply

Captures from 1 September 2026 record a category-wide move to consumption units. Zapier has unified its pricing so that AI steps, code and SDK all follow the same task-based model, with usage measured in tasks across all features and tiers running from 100 to two million tasks.

n8n lists Starter at €20/month billed annually for 2,500 executions and 2,300 AI credits, Pro at €50/month for 10,000 executions and up to 13,700 AI credits, and Business at €667/month, with a Start-up plan at 50% off Business for companies under 20 employees. The Community Edition remains available, and the free trial gives Pro features at Starter limits.

UiPath starts at $25/month for a Basic plan aimed at individuals and small teams, with Standard and Enterprise requiring a sales conversation. Lovable prices in credits outright and — unusually, and to its credit — publishes example costs: simple UI changes at 0.5 to 0.9 credits, adding authentication at 1.2 credits, building a landing page with images at 1.7 credits, and one credit per message in Plan Mode.

A seat price answers "what will this cost?" A credit balance answers "what will this cost if nothing changes?" Those are very different promises, and the second one is the one you are now being sold.

The risk is not that credits are expensive. Often they are cheap. The risk is that consumption is a function of how much your team uses the tool, and successful adoption of an automation platform looks exactly like a cost overrun until you can attribute it.

This is where Lovable's published per-action costs earn a mention. Knowing that a landing page costs about 1.7 credits lets you forecast. A vendor that gives you a credit balance and no per-action figures has handed you a budget you cannot build.

Pick three workflows you actually run. A form-to-CRM sync. A document summarisation job. An approval routing flow with an AI classification step.

For each platform you are considering, work out what one run of each workflow consumes in that platform's unit — tasks for Zapier, executions plus AI credits for n8n, credits for Lovable — and multiply by your real monthly volume. Only then look at the plan tiers.

You will usually find the ranking changes. A platform with a higher plan price and a generous unit can be cheaper than a cheap plan that charges a task per step in a ten-step workflow. This exercise takes an afternoon and is the single most useful thing on this page.

Notice that n8n prices executions and AI credits as separate quantities. That separation is honest, and it points at something worth acting on: the automation is cheap and the model call inside it is not.

If a workflow's AI step is a simple classification or extraction, ask whether it needs to run through the platform's credit system at all. Calling a cheap model directly and passing the result back into the workflow can cut the AI portion substantially, at the cost of a little more plumbing.

That trade is not always worth it. A platform-managed AI step is easier to maintain and easier for non-engineers to change. But you should make that trade knowingly, and right now most teams are making it by default because the AI step was the easiest box to drag onto the canvas.

Seat-based governance asks "who has a licence?" Consumption-based governance has to ask "what is running, how often, and who owns it?"

Three rules cover most of it. Every scheduled automation has a named owner. Every automation has a documented expected monthly run count, checked against actual. Any automation that can trigger itself in a loop has a hard cap.

That third rule is not theoretical. The cheapest way to discover consumption pricing is a recursive workflow that ran forty thousand times over a weekend, and the fix costs nothing to put in place beforehand.

Best for predictable, knowable volume: an execution-priced plan such as n8n Starter at €20/month annually for 2,500 executions, or the Community Edition if you can host it.

Best for broad, non-technical adoption across a business: Zapier's unified task pricing, precisely because one unit across AI, code and SDK steps makes the bill explicable to people who will never read a pricing page.

Best for build-style AI work: Lovable, on the strength of published per-action credit costs — but forecast at your real prompt volume, because 1.7 credits per landing page adds up faster than it reads.

Avoid: comparing monthly plan prices across these vendors without normalising the unit first, and running any self-triggering automation without a cap. Confirm current figures on each vendor's pricing page — these were captured on 1 September 2026 and this category moves.