The Short Answer
Zapier charges by the task — a unit of completed work, not a flat subscription for whatever you build. Its own pricing page defines a task as any Zap step, AI step or code step that completes successfully; triggers, polling and the built-in logic tools (Filter, Paths, Formatter, Delay, Looping) cost nothing at all. That sounds simple until a workflow starts running several action, AI or code steps for a single business event, because Zapier's own documentation says plainly that "many actions cost more than one task — the amount depends on step type, AI model tier, and number of tool calls." The bill is not one task per event. It is one task per successful step, and a genuinely useful automation is rarely one step.
A custom AI automation build works differently: a one-off engineering cost, then a flat monthly running cost — model usage and hosting — that does not climb because a workflow ran more steps or handled more events this month.
Neither is the wrong answer by default. This guide works from Zapier's own published pricing structure (checked directly against zapier.com/pricing and Zapier's help centre, both current as of September 2026) rather than a competitor's marketing about it, so you can see which side of the trade-off your own workflow sits on.
What Actually Counts as a Task
This is the part worth reading slowly, because it decides your real running cost far more than the headline plan price does.
Zapier's own pricing reference states that "a task is counted whenever Zapier successfully completes a unit of work for you. Failed actions are not counted." Specifically metered: a successful standard Zap step, an AI by Zapier step, and a Code by Zapier step. Not metered at all: the trigger that starts a Zap, polling for new trigger data, and Zapier's built-in logic tools — Formatter, Paths, Filter, Delay, Looping, Sub-Zap, Digest, Storage and Tables.
That free logic layer is genuinely generous — you can branch, filter and reshape data for nothing. What is not free, and what the pricing page states outright, is that "many actions cost more than one task — the amount depends on step type, AI model tier, and number of tool calls." An AI step that makes several tool calls to complete one job, or a multi-step Zap that fires four action steps off one trigger, can consume several tasks for what looks to the business like a single event. Nobody sizing a plan against "how many orders we get a month" is sizing it against the right number.
Running out matters more than it sounds. Cross a plan's task limit and Zapier automatically switches you to pay-per-task billing "so your Zap workflows stay on," at a published overage rate of 2.5× your base rate on monthly billing, or 1.25× on annual. That is capped, not open-ended: usage stops once it reaches three times your plan's task allocation, until the next billing cycle. You can turn pay-per-task off, but the alternative is your workflows simply stopping when the free allocation runs out.
Where Zapier's Own Free Tier Actually Stops
The Free plan is not a smaller version of Professional — several things that matter for a real workflow are gated to paid plans entirely, per Zapier's own feature comparison:
| Capability | Free plan | Paid plans (Professional and above) |
|---|---|---|
| Zap steps | Two-step only (one trigger, one action) | Multi-step — one trigger, multiple actions |
| Filters and Paths (branching logic) | Not included | Included |
| Formatter (reshaping data between steps) | Not included | Included |
| Premium app connections | Not included | Included |
| Webhooks (custom integrations) | Not included | Included |
| AI by Zapier steps | Not included | Included |
| Code by Zapier runtime per step | 1 second included; extended runtime not available | 30 seconds included (Professional/Team), 2 minutes (Enterprise); extended runtime available as an add-on |
That last row is the one most comparisons skip. Zapier itself ships a code step for JavaScript and Python — tacit agreement that no-code logic eventually needs real code — but it is metered by the second. Runtime beyond what your plan includes is rounded up to the next 30-second block, and Zapier's own documentation states each block "uses 1 task." A workflow with genuinely complex logic can spend its task budget on the code step alone.
Where Zapier Genuinely Wins
We connect client systems through Zapier, Make and n8n ourselves when it is the right tool — the orchestration layer is rarely what a project actually needs to get right; what each business system permits is. Zapier earns its place when:
- •The logic is simple and the volume is low. A single trigger firing a single action, a handful of times a day, is exactly what the free logic tools and a modest task allowance are built for.
- •You need it live this week. Point-and-click setup against a library of pre-built app connections beats a build timeline every time, if the connection you need is already there.
- •Nobody owns custom software internally. A platform is maintained centrally by its vendor. A custom build needs someone — us or you — to look after it.
- •The workflow will not grow much. If this is genuinely a one-off, steady-state task, the task ceiling and the per-step metering rarely become a problem worth solving.
Where a Custom Build Wins
- •The event triggers several steps, not one. Because every successful action, AI and code step is its own metered task, a workflow that does five things per order or per enquiry is paying five times over on a per-event basis — a custom build's running cost does not multiply the same way.
- •The logic outgrows Filters and Paths. Real business rules — the exceptions that only live in someone's head, decision trees with several conditions, logic that changes based on data from more than one system at once — get harder to express cleanly in a visual builder long before they get impossible, and every extra step to express them is another metered task.
- •You need genuine code, not a 30-second block. Zapier's own Code by Zapier runtime caps (1 second on Free, 30 seconds on Professional/Team, 2 minutes on Enterprise) are workable for a short transform, not for anything computationally real. A custom build has no per-step runtime ceiling.
- •You want the asset, not a subscription to it. A custom build's logic, prompts and integrations are yours. They keep working if a vendor changes a pricing structure, a task definition, or an overage rate — which Zapier, like any SaaS platform, can and does revise.
- •Approval and audit steps matter. Our own automation builds put a human approval step in front of anything that touches money or a customer record, designed in from the start rather than bolted on — see how that works on [/services/ai-automation](AI Automation).
Working Out Which Side You're On
You do not need Zapier's exact rate card to know which way this leans for you — you need your own event volume and how many steps each event genuinely requires.
| Your workflow | What happens on Zapier | What happens with a custom build | Generally favours |
|---|---|---|---|
| Low volume, one trigger → one action | Comfortably inside a modest plan; free tools cover simple branching | Build cost has little volume to spread across | Zapier |
| Growing volume, several action or AI steps per event | Task usage climbs faster than event volume, because each step is its own task and AI steps can cost more than one | Running cost stays flat — hosting and model usage, not per-step metering | Roughly level, tipping toward custom as steps-per-event rises |
| High volume, or logic that needs real code and cross-system decisions | Code by Zapier's per-step runtime caps and the pay-per-task overage (up to 3× your plan allocation) become a recurring cost of their own | No per-step runtime ceiling, no per-action task meter | Custom, usually by a wide margin |
Questions to Settle Before You Choose
- 1. How many action, AI or code steps does one business event actually trigger — not "how many events a month," but steps per event?
- 2. Have you hit your task limit and switched to pay-per-task billing already, even once?
- 3. Does the logic still fit cleanly in Filters and Paths, or are you fighting the visual builder?
- 4. Would a Code by Zapier step need more than 30 seconds to run, even occasionally?
- 5. Do you want to own the automation's logic outright, or is a subscription to someone else's platform fine?
If your answers point at a simple, low-step, steady workflow, Zapier is very likely the right tool — including built by us, if you want it done properly rather than assembled by trial and error. If they point at multi-step logic, real code, or task costs that already climb faster than your business does, a custom build is worth pricing against what Zapier is actually costing you today, overage included.
How We Approach It
We do not default to "fully custom" as a sales position. On the discovery call we look at what you have built or want to build in Zapier, Make or n8n first, and we will tell you plainly if the platform you already pay for is the right tool — sometimes the honest fix is restructuring the Zap, not replacing it. Where the task-based model genuinely stops fitting — the step count, the runtime cap, or the logic itself — we build the workflow directly, connected to your own systems on your own accounts, with the human approval step designed in from the start.
The systems we connect to and how a build is scoped week by week are on [/services/ai-automation](AI Automation). For a sector where this trade-off comes up constantly — invoice processing, client onboarding, practice management sync — see [/industries/accountants](AI for Accountancy Firms). If you are further back and still weighing what an agency actually charges for automation work when the builder tools themselves are nearly free, [/blog/ai-automation-agency-cost-uk](What an AI Automation Agency Costs in the UK, and What You Are Paying For) covers that directly, and [/blog/how-long-does-an-ai-chatbot-take-to-build-uk](How Long Does an AI Chatbot or Automation Actually Take to Build in the UK?) sets out realistic timelines across all four AI services including automation.
Want your own task usage and step count looked at against a fixed-cost build? Book a free discovery call at [/contact](our contact page).
Frequently Asked Questions
Does Zapier already do most of what a custom automation build does?
For a simple, low-step workflow, often yes — Zapier’s free logic tools (Filter, Paths, Formatter) cost nothing to use, and its library of pre-built app connections covers most common business systems out of the box. Where it stops fitting is volume and complexity: every successful action, AI or code step is its own metered task, so a workflow that fires several steps per event is paying per step, not per event, and that adds up in a way a flat-cost custom build does not.
How does Zapier’s task-based pricing actually work?
Per Zapier’s own pricing documentation, a task is counted whenever a Zap step, AI step or code step completes successfully — triggers, polling and built-in logic tools are free. Cross your plan’s task limit and you are automatically moved to pay-per-task billing so workflows keep running, at a published overage rate of 2.5× your base rate monthly or 1.25× annually, capped at three times your plan’s allocation per cycle. Many individual actions cost more than one task depending on the step type and, for AI steps, the model tier.
Can Zapier run our own custom logic instead of a full custom build?
To a point. Code by Zapier runs JavaScript or Python on every plan, but the included runtime per step is capped — 1 second on Free, 30 seconds on Professional and Team, 2 minutes on Enterprise — and anything beyond that is billed in 30-second blocks, each costing an additional task. That is workable for a short data transform. It is not a substitute for genuine custom logic that needs real processing time.
What is the actual signal that a business has outgrown Zapier?
Repeatedly hitting the task ceiling and relying on pay-per-task overage as a normal monthly cost, workflows that need more branching or cross-system decisions than Filters and Paths express cleanly, or code steps that regularly need more runtime than the plan includes. Any one of those means the task-based model is now working against you rather than for you, and it is worth pricing what Zapier currently costs at your real volume against a fixed-cost build.
Do you build fully custom, or do you build on top of Zapier and similar tools?
Both, depending on what the workflow actually needs — we connect client systems through Zapier, Make and n8n ourselves where that is the right, fastest and cheapest tool, and we will say so on the discovery call rather than defaulting to a bespoke build. Where the task-based model, the step count or the logic itself stops fitting is when we build the automation directly, on your own accounts, with a flat running cost and a human approval step designed in from the start.