What a zap bills for, and what a run covers
Most pages about a Zapier alternative compare app counts and plan tiers. The more useful comparison is the unit you pay for. Zapier bills per task: most steps in a zap consume a task each time it fires, so the same workflow costs more as it grows longer. A run is one billed execution of a playbook. The research, the drafting, and the formatting inside it are one price, however many steps the agent takes to finish the job.
The failure story differs too. In Moatt, a run that fails refunds its credits automatically, so you are never billed for a job that did not complete. Talking to the agent costs nothing: planning, questions, and back-and-forth are free. Credits only move when a playbook actually executes, and you see the price at the approval gate before it does.
Per-task pricing vs per-run credits
Per-task pricing grows on two axes at once: how often your automations fire and how many steps each one has. Add a filter or a formatting step and the same monthly volume costs more. Past your limit, Zapier moves usage to pay-as-you-go unless you turn that off. None of this is wrong. It just means your bill is a function of workflow shape, not of what the work was worth.
Credits price the job instead. One credit is one standard run, and the price shows at the approval gate before anything fires. From the rate card:
The guardrails are fixed rules, not fine print. Unused credits roll over up to twice your monthly amount. Overage bills at 1.25x the base rate and caps at 3x your plan, so a busy month cannot produce a surprise invoice. Alerts fire at 50, 80, and 100 percent of usage.
- Monitoring sweep — 1 credit
- Reply or outreach batch (5 drafts) — 1 credit
- Keyword gap report — 2 credits
- Page written and formatted — 3 credits
- Competitor profile — 3 credits
The approval gate changes what you can automate
A zap fires the moment its trigger does. If you want a human check, you build it yourself with a delay, a review table, another zap. In Moatt the gate is the default: nothing sends, posts, or publishes without your approval. A run finishes its work, then waits with the drafts and the price attached. You approve, edit, or kill it.
That one mechanic changes what is safe to automate. Nobody lets a trigger-action tool answer prospects unattended, so the risky-but-valuable jobs stay manual. With a gate, the drafting and research get automated while the judgment call stays yours. Scheduled runs work the same way: one sentence in chat becomes an automation that fires with your laptop closed, and anything outbound still stops at the gate.
Which zaps to move to an AI agent first
Start with the zaps that end in a person reading something, because those are the ones where the tool stops short of the actual job:
Leave the pure data syncs alone at first. The order matters because runs feed memory: outcomes come back as lessons backed by measured results, kept apart from untested hypotheses. The tenth monitoring sweep knows what the first nine learned. A zap runs the same on day one and day three hundred — that is its strength for plumbing and its ceiling for growth work.
- A mentions-to-Slack pipe becomes a monitoring sweep (1 credit) that scores what it found and drafts the replies worth sending.
- A new-lead alert becomes an outreach batch: five drafts waiting at the gate for 1 credit.
- A content chain stitched across three tools becomes a keyword gap report (2 credits) that turns into a written, formatted page (3 credits) in the same thread.
When Zapier stays the right tool
Some jobs should not move. If the mapping between trigger and action never needs a decision — a form submission becomes a CRM row, an invoice copies to a folder — Zapier does it cheaply and instantly, and the niche tool you use is probably already supported. Jobs that must run fully unattended belong there too: Moatt's approval gate is the point for growth work and pure friction for plumbing. Plenty of teams run both, and should.
A workable rule: if you could write the step in the middle as a spreadsheet formula, keep it as a zap. If the step in the middle is "a person decides," it is a run. Moving the second kind first is where the switch pays for itself.