AI Automation Cost: 3 Pricing Models Compared

A client emailed me last month: “I want to automate our lead intake. What does that actually cost?” She had three quotes. A no-code agency wanted $400 a month on Zapier plus $2,000 setup. A boutique dev shop wanted $18,000 for a custom build. Someone on Upwork offered a “ready-made agent” for $250 flat. All three pitched the same outcome: inbound form to CRM to Slack to a first-touch email. The prices spanned two orders of magnitude. She couldn’t tell which was fair.

She isn’t unusual. Most small operators walk in with no map of the pricing terrain, so the first quote becomes their anchor. That’s how you end up on a $600/month Zapier plan for a workflow that could have been a $150 Python script, or paying $25K for a custom build when a $50 template pack would have done the job. This post is the map: three tiers, honest USD ranges, a decision tree at the end. I run Solo Stack Studio as a one-person operation with eight AI departments, and I’ve built or bought inside all three tiers over the last two years.

Tier 1: Subscription — pay monthly, own nothing

The entry point. You sign up for Zapier, Make.com, n8n cloud, or OpenAI Assistants API, connect a few apps, and pay a recurring fee. You can be running in a weekend.

Pricing as of mid-2026:

• Zapier: Starter $30/month (750 tasks), Professional $75-100/month (2K tasks), Team $400-800/month (50K+ tasks). Tasks stack fast with multi-step Zaps, filters, and formatters.
• Make.com: Core $10/month (10K ops), Pro $16-50/month, Teams $100-400/month. Cheaper per operation than Zapier, steeper learning curve.
• n8n cloud: Starter $20/month, Pro $50/month, Business $120-200/month. Self-hostable if you want to escape the meter.
• OpenAI Assistants API: $20-200/month for a small business use case, depending on token volume and whether you use GPT-4-class models. Add vector store fees for retrieval.

The hidden cost is task inflation. A workflow that costs 3 tasks per lead today costs 8 six months later, once you add “check for duplicates,” “enrich with Clearbit,” “route by territory,” and “send follow-up sequence.” I’ve seen a client’s Zapier bill go from $75 to $520 over eighteen months without a single new automation added. Same workflow, doing more per run.

Best for: Quick wins where the workflow is stable and low-volume. A founder testing an idea. Fewer than five automations, none running more than a few thousand times a month, and ownership doesn’t matter — you’ll rebuild if it works.

Not for: Anything running 50,000+ times a month (the pricing curve becomes brutal), custom logic beyond what the visual builder offers, or anything you’d want to migrate later. Subscription tools lock your logic inside their UI. Migration is a rebuild.

Tier 2: One-time buy — templates, skill packs, prebuilt agents

The tier most people don’t know exists. You pay once for a working artifact — a Make.com scenario template, a Zapier Zap pack, a Claude skill, an n8n workflow file — and own it forever. No recurring fees except whatever platform you run it on.

Real pricing:

• Skill Store agents / Claude skills: $100-500 per skill on marketplaces like PromptBase, or bundled at $50-200 in template packs. My claude-code-skill-stack is free; the same shape sells for money elsewhere.
• n8n / Make template packs: $50-300 for a bundle of 10-30 workflows around a theme (lead gen, ecommerce ops, content publishing). Zapier Templates, Make’s public library, and n8n.io have free ones too.
• Prebuilt solutions: $300-2,000 for an “install this and it works” package with documentation, video walkthrough, sometimes 30 days of email support. Common for real estate CRMs, Shopify order flows, and podcast production pipelines.

The math flips fast. A $200 skill pack that replaces a $75/month Zapier plan pays back in under three months and keeps paying after that. The tradeoff: you install it, adapt it to your data, and maintain it when a downstream API changes. If you can read documentation and edit YAML, this tier is dramatically underpriced.

Best for: Repeatable tasks someone else already solved. You want ownership so you can modify without asking permission, and you’re okay reading a README and running a few CLI commands. High-volume workflows where per-task pricing would kill you.

Not for: Novel workflows nobody has built before, deep integration with a proprietary internal system, or teams that need a vendor relationship for compliance reasons.

Tier 3: Custom build — hire someone to build it for you

You pay a developer, agency, or consultant to build from scratch to your specification. Prices scatter wildly because “custom” covers everything from a weekend script to a six-month enterprise integration.

Honest ranges based on what I see in the market and what I quote myself:

• Small custom ($3K-8K): One workflow, 2-3 integrations, standard logic. Two to four weeks. Fits “automate this one painful process end to end.” Delivered as code you own, deployed to a platform you control.
• Mid custom ($10K-25K): Multiple workflows, 5+ integrations, some custom logic (scoring, routing, decisioning). One to three months. Fits “replace this team’s manual work” or “grow 3x without hiring.”
• Large custom ($30K+): Cross-departmental system, custom UI, ongoing maintenance. Three to six months plus a $2K-5K/month retainer. Fits “this is core infrastructure and we need it right.”

What you’re paying for isn’t the code. Modern AI tooling means an experienced builder can produce a working prototype in a day. You’re paying for discovery: the four hours of interviews separating what actually needs automating from what people say they want. You’re paying for a migration path, for the second draft after the first one hits reality, and for someone still there in six months when an API deprecates and something breaks.

The failure mode is scope. I’ve watched clients spend $40K on custom builds that could have been $200 template packs, because the vendor never mentioned the alternative. I’ve also watched clients spend $2K on a template pack for a workflow that genuinely needed $15K of custom work, then spend six months bandaging the mismatch. Good consultants tell you which tier you need. Bad ones sell the tier with the best margin.

Best for: Unique workflows with no off-the-shelf equivalent, cross-departmental systems, workflows that touch sensitive data or need audit trails, or volume so high that per-task subscription pricing would be prohibitive. Also when you need someone accountable when things break.

Not for: Standard SaaS integrations that have a template pack, unvalidated exploratory ideas, or teams without an internal owner to maintain the system after handoff.

Decision tree: which tier are you actually in?

Skip the vendor pitches and answer these in order.

Question 1: Standard or novel? If someone already sells a template for it (lead intake, invoice processing, content scheduling, e-commerce order flow), start at Tier 2. If it’s genuinely unique, skip to Question 3.

Question 2: Will you install a template yourself? If yes, Tier 2 is dramatically the best value. If no, and the workflow is standard, Tier 1 for the first six months while you validate, then reevaluate: template or build.

Question 3: What’s the monthly volume? Under 5,000 runs a month, Tier 1 is fine even for custom-feeling workflows. 5,000-50,000, Tier 2 templates or Tier 1 on Make/n8n. Over 50,000, custom build or self-hosted n8n. The subscription math stops working somewhere between 20K and 100K runs depending on step count.

Question 4: How critical is ownership? If you need to audit the code, comply with a data regulation, or survive your vendor folding, Tier 2 or Tier 3. Tier 1 gives you no ownership of logic.

Question 5: Are people in the loop? If the automation coordinates with humans (approvals, handoffs, exception handling), you almost always end up in Tier 3. Too many edge cases for templates.

Most solo operators are in Tier 2 territory and don’t know it. Most small teams (5-20 people) start in Tier 1, then move to Tier 2 or Tier 3 by their fifth automation. Enterprise starts at Tier 3.

What I’d do in your position

If you’re just starting, spend $200 on a Tier 2 template pack that covers your most painful workflow. Install it yourself, break it, fix it, learn how the pieces connect. That single weekend teaches you more than three months of researching pricing pages.

Once you know what “good” looks like, you can make honest tier decisions for the next five workflows. Some will stay in Tier 1 forever because the volume is too low to justify anything else. Some will jump straight to Tier 3 because they’re mission-critical from day one. Most will live in Tier 2.

For the full framework on scoring which workflows to automate first and building a stack that grows with you, read Solo Stack Method. Weighing a specific custom build? See the consulting framework and ROI calculator. The worst outcome is not paying too much or too little. It’s building the wrong thing at any price.

FAQ

How do I know if my Zapier bill is bloated?

Divide monthly cost by automations you actually use. Above $30 per active automation, you’re probably paying for tasks you don’t need. Check the task history for your top three Zaps. If one Zap uses more than 40% of your monthly tasks, that’s your candidate to migrate to a script or template.

Can I move a Zapier workflow to n8n later?

Not automatically. You rebuild the logic in n8n’s visual editor or as code, 2-4 hours per Zap depending on complexity. This is why locking mission-critical logic into a subscription tool is risky: the migration cost is the exit tax.

What’s the payback period on a custom build?

A $10K build replacing 10 hours a week of manual work at a $50/hour internal cost breaks even in about five months. A $25K build enabling revenue you couldn’t have captured otherwise (faster response times, larger deal capacity) often pays back in the 6-12 month range depending on deal size. Longer than 18 months means the ROI case is weak. Reconsider whether you need custom.

Should I hire a consultant or a developer?

Consultants scope the problem. Developers build the solution. For a $3K-8K build, one senior developer who does discovery well is enough. For $15K+, you want someone who scopes before writing code, usually a consultant with implementation capability rather than a pure dev shop.