Automation Retainer Pricing: Stop Selling One-Off Builds

Automation retainer pricing comes down to one decision: are you selling a build, or a system that keeps working? Sell the system. Charge a setup fee priced on the value of the problem, a monthly retainer that pays you to keep it running, and pass platform and AI usage through to the client at cost or close to it. Hourly billing and build-only projects are how automation freelancers end up doing free maintenance for clients they closed a year ago.

I build a white-label automation platform for agencies, so I spend a lot of time with people running a handful of client accounts on Make, n8n or Zapier. The ones who are burned out almost always sold builds. The ones with calm months sold retainers and wrote down exactly what the retainer covers. This is how I’d price, package and sell automation work if I were starting over.

How should you price automation services?

Price automation services as three separate lines: a one-time setup fee tied to the value of the problem you’re solving, a flat monthly retainer for keeping the automation working and improving, and platform and AI usage passed through transparently. Don’t sell hours, and don’t bury variable usage inside a flat fee.

The hourly market is crowded and cheap. Upwork puts freelance n8n experts at roughly $40 to $100 an hour[1] and Zapier developers at $40 to $90[2]. On Contra, n8n freelancers list at $15 to $30 an hour at entry level and $50 to $80 or more at expert level[3]. Quote by the hour and that’s the shelf your client compares you to.

Hourly billing also punishes you for getting good. The third lead-routing workflow you build takes a fraction of the time the first one did, because you already solved the edge cases. Bill it by the hour and your reward for expertise is a smaller invoice.

Price the setup on the problem, not your time

Start with what the manual process costs today. Say the client’s office manager spends 10 hours a week copying web leads into the CRM and texting the sales team. At a loaded cost of $35 an hour, that’s $18,200 a year. A $4,000 setup fee pays back in under three months, and the client can check that math on one line. The same build billed at $80 an hour for 15 hours comes to $1,200. Same outcome, $2,800 left on the table.

Most agencies still don’t price this way. In SoDA and Productive’s agency survey, value-based fees made up only about a tenth of agency revenue[4], which is exactly why pricing on outcomes still stands out in a sales call. The creators teaching this crowd land in a similar place. Nick Saraev suggests selling common automation systems for roughly $1,500 to $5,000 fixed price, with a maintenance or management fee on top[5]. A 2026 rate card from Bet on AI, built from invoices across 54 freelancers and agencies, put a lead-capture-to-CRM build at $650 to $1,400 on Zapier or Make and a sales pipeline with AI scoring at $3,500 to $6,500[6]. It’s a small sample with a light methodology, so I’d use those numbers as floors to anchor against, not targets.

Automation retainer pricing: setup fee vs retainer vs usage

Each line pays for something different, which is why they shouldn’t blur together. The setup fee pays for discovery, the build, testing and handover. The retainer pays for every month after launch, when things break or change. Usage covers the vendors. Blend them and you end up absorbing your client’s growth as your own cost.

• Setup fee, one time. Covers mapping the process, building, testing against real client data, and a written runbook. I’d take half up front and half at go-live.
• Monthly retainer. Monitoring, fixes when something upstream changes, a capped allowance of small changes, and a monthly report. Set a three-month minimum, since the first month after launch is always the noisiest.
• Usage, passed through. The automation platform plan and AI model tokens. The client pays these at cost, either in their own accounts or as a separate line on your invoice.

Running more than one pricing model is normal. Promethean Research’s industry report, drawn from 1,452 agency leaders, found that 8% or fewer agencies rely on any single model[7]. What matters is that each line has one job and the client can see it.

Why build-only pricing loses money

Build-only pricing assumes an automation is finished when it ships. It isn’t. APIs version and retire, AI models get deprecated on a schedule, and clients rename the CRM field your workflow reads. Somebody has to fix that. If nobody is paying for it, the somebody is you working for free, or nobody, and the client blames you anyway.

The churn runs on a published calendar. Meta stops serving each Graph API version two years after the next one is released[8], and it shipped five versions between January 2025 and July 2026[9]. Anthropic retired nine Claude model versions between October 2025 and August 2026, with at least 60 days’ notice each time[10]. OpenAI shut down its Assistants API on August 26, 2026, a year after announcing it[11]. Each of those dates meant workflows somewhere stopped working until a person opened them and made an edit.

A retainer is how you get paid for that edit. It also turns a breakage from a complaint into evidence that the retainer is worth paying for.

What an automation retainer should include

A good automation retainer covers four things in writing: monitoring with a stated response time, fixes when anything upstream changes, a capped allowance of small changes each month, and a short monthly report. It should also say plainly what it excludes. The big one is new workflows, which get quoted separately with their own setup fee.

• Monitoring with a response time. You watch every run, and failures reach you before the client notices. Put a number on it: next business day for routine failures, four business hours for anything touching revenue.
• Breakage fixes, uncapped within scope. An API change, an expired token or a retired model on an existing workflow is your problem, included. This promise justifies the fee, so don’t meter it.
• A change allowance. Three small changes or four hours a month, whichever you’d rather track. Define small in the contract (a new field, a reworded message, an extra filter) so a new trigger or a new app is obviously a new project.
• Connection upkeep. Reconnecting expired OAuth tokens and chasing the client when a connection needs their login.
• A one-page monthly report. Runs, failures, what you did about them, and the hours or dollars the automations saved.

If I had to cut something, I’d cut the change allowance before the report. The report is the only moment the client actually sees the work, and it’s what renews the retainer.

What to charge for the retainer

Solo operators I talk to mostly land between $500 and $3,000 per client per month, and the Bet on AI rate card lines up with that: $650 to $1,200 for a care tier with monitoring and one small build a month, $1,200 to $2,400 for two or three workflows a month, and $2,400 to $3,800 for a growth tier with a weekly review and AI tuning[6]. My rule of thumb is a monthly fee of 15% to 25% of the setup fee, and never less than what one bad week of breakage would cost you in unpaid hours.

How to handle platform and AI costs

Pass platform and AI usage through to the client, either in accounts the client owns or as a separate invoice line at cost plus a small handling fee. Never fold variable usage into a flat retainer. Usage grows with the client’s volume, and a flat fee means their growth comes straight out of your margin.

Whose name the platform account sits under is its own decision, with real consequences when a client leaves, and I wrote that up in who should own client automations. For pricing, the rule holds either way: the client should be able to see what the vendors cost. The billing unit matters too. Per-task and per-credit plans multiply with steps and records, so a client who doubles their lead volume can double the bill. I compared the units in automation pricing models compared.

AI is the line that surprises people. Token spend moves with prompt length, retries and how chatty an agent loop gets. The Bet on AI rate card reports that LLM costs are almost always passed through at cost plus a 10% to 15% management fee[6], and that’s what I’d do, with two guardrails. Agree a monthly AI spend cap per client up front, and keep a separate key per client, whether in their name or yours, so the invoice maps cleanly to the usage. There’s more on key custody in BYOK for automation agencies.

How to sell an automation retainer on a sales call

Sell the retainer in the first proposal, never as an upsell after launch. Price the problem first, then present one package: setup fee, retainer, and a usage estimate. Explain the retainer by naming what will break in their stack, and give the client an honest, more expensive alternative if they decline it.

Spend the first half of the call on what the manual process costs. Hours per week, who does them, what a dropped lead or a late invoice costs. Write the annual number where they can see it. Everything you quote afterward gets compared to that number instead of a $40-an-hour marketplace profile.

Put all three lines on one page. Clients rarely balk at a retainer they saw on day one; they balk at one that appears in month two. For anything bigger than a single workflow, sell a paid discovery first, a fixed-price process map and build plan. It filters out people who only wanted free consulting.

Make the retainer concrete with their own tools. If they run Facebook lead ads into HubSpot, tell them Meta retires API versions on a fixed schedule and the upgrade is on you. That lands harder than any generic line about peace of mind.

And when a client says they don’t want the retainer, don’t discount it away. Offer a clean handover instead: documentation, accounts and credentials in their name, and future fixes billed hourly at a rate above your retainer’s effective rate. Priced honestly, the retainer is usually the obvious choice.

One more thing: keep the menu short. Promethean Research found that agencies that cut back their service list grew 13% on average and posted 30% net margins, against a 13% average net margin across the sample[12]. Three packaged retainer tiers beat a custom quote for every prospect.

Frequently asked questions

How much should I charge for an automation retainer?

For a small business with a handful of workflows, $500 to $1,500 a month is a common range, rising to $2,400 to $4,000 when the retainer includes new builds every month or active AI workflows. Anchor it to the value the automations protect, and make sure it covers the unpaid hours a bad month of breakage would otherwise cost you.

Should I charge hourly for n8n or Make work?

Only for small, ad-hoc jobs outside a retainer. Hourly pricing puts you next to $40-an-hour marketplace profiles and pays you less as you get faster. A fixed setup fee plus a retainer lets you keep the benefit of the patterns you’ve already built.

What happens when a client cancels the retainer?

Hand over cleanly. Confirm the platform accounts and credentials sit in the client’s name, deliver the runbook, and agree an hourly rate for future fixes. Cancelling should end your obligation to maintain the automation, not the client’s ability to run it.

Should I mark up AI and platform costs?

A 10% to 15% handling fee on usage you pay for and rebill is normal, since you carry the cash flow and the reconciliation. Don’t hide a bigger margin in usage. Put your margin in the retainer, where the client can see what it buys.

Once you run retainers for more than a few clients, those promises become operations work: who sees a failure first, where change requests get logged, what the monthly report pulls from. I wrote that playbook in managing client automations. We built TaskJuice around this model. Each client gets an isolated workspace with its own run history and a white-labeled portal where they see their runs and submit change requests that you quote and they approve, which maps neatly onto a retainer’s change allowance. You can bill the retainer and one-off builds through Stripe Connect, and set a monthly AI spend cap per client workspace so pass-through costs stay inside what you agreed. Whatever you run it on, stop selling builds. Sell a system that keeps working, and get paid for keeping it that way.

References

[1] Upwork: Hire n8n Experts: www.upwork.com/hire/n8n-experts/

[2] Upwork: Hire Zapier Developers: www.upwork.com/hire/zapier-developers/

[3] Contra: Hire n8n Freelancers: contra.com/hire/n8n-freelancers

[4] Productive: In Pursuit of the Best Agency Pricing Model (SoDA and Productive survey): productive.io/blog/best-agency-pricing-model/

[5] Nick Saraev: 5 More Automations You Can Sell Today for $1,500 (Or $10,000): nicksaraev.com/5-more-automations-you-can-sell-today-for-1-500-or-10-000/

[6] Bet on AI: AI Automation Rate Card 2026, Real Rates From 54 Operators: betonai.net/ai-automation-rate-card-2026-what-to-charge-for-n8n-make-and-zapier-builds-real-rates-from-54-operators/

[7] Promethean Research: Digital Agency Industry Report: prometheanresearch.com/digital-agency-industry-report/

[8] Meta for Developers: Graph API Versioning: developers.facebook.com/docs/graph-api/guides/versioning

[9] Meta for Developers: Graph API Changelog, Versions: developers.facebook.com/docs/graph-api/changelog/versions

[10] Anthropic: Model Deprecations: platform.claude.com/docs/en/about-claude/model-deprecations

[11] OpenAI: Deprecations: developers.openai.com/api/docs/deprecations

[12] Promethean Research: 2026 State of Digital Services: prometheanresearch.com/2026-state-of-digital-services-digital-agency-industry-research/