AI automation builds are the highest-margin AI consulting offer in 2026: a single workflow — lead enrichment, invoice processing, AI content pipeline — bills $1,500–$5,000 for the build and $500–$1,500/mo to manage. The clients aren't technical. They're business owners who know something is broken and want it fixed, not another SaaS license.
The stack that closes is Gumloop for AI-native workflows, n8n for clients who want to own the infrastructure, and Make for stitching SaaS apps together. Claude writes the logic; Zapier handles the long tail of smaller integrations. Total tool cost: under $70/mo on a single client engagement. First build produces its own case study.
The tools
Gumloop
Gumloop is built for AI-native workflows — chain LLM steps like scraping, extraction, classification, and generation into a visual canvas without code. Built-in access to multiple AI models means no separate API keys, and its credit-based pricing stays flat whether your clients' workflows run 10 times or 10,000 times a month. It's the fastest path from "I need AI in my process" to a delivered workflow.
Price: Free (5,000 credits/mo, 1 seat); Pro $37/mo (20,000+ credits/mo, unlimited seats; ~$29.60/mo annual). What to charge: $1,500–$3,500 per workflow build, $250–$500/mo for credit monitoring and prompt tuning. Frame it as an AI implementation, not automation — that framing supports 2× the price of equivalent Zapier work.
n8n
n8n is the right tool for clients who want to own their automation infrastructure. Self-host it on a $20/mo VPS (Community Edition is free, unlimited executions) and bill a $500–$1,500/mo managed service retainer with 90%+ margin. The execution-based pricing model — one run costs the same whether it has 3 steps or 300 — changes the economics on heavy AI-agent workflows.
Price: Community Edition self-hosted free; Cloud Starter €20/mo annual (€24/mo monthly, 2,500 executions/mo). What to charge: $2,500–$5,000 per build, then $500–$1,500/mo for managed hosting, monitoring, and changes.
Make
Make is the most forgiving entry point for clients already running SaaS apps (HubSpot, Notion, Airtable, QuickBooks). The visual scenario builder and 2,000+ app integrations handle the connective tissue while your AI layers sit on top. The free plan (1,000 credits/mo) covers your own prototyping before you bill anything.
Price: Free (1,000 credits/mo); Core $9/mo annual, Pro $16/mo annual. What to charge: $1,500–$3,000 per multi-app automation system, $300–$500/mo to maintain and expand.
Claude
Claude is the reasoning engine inside every build: it drafts the workflow logic, writes prompt nodes, generates the output templates, and — critically — it's the tool you use to interview the client about what's actually broken before you touch a canvas. A Client Project in Claude holds every document, so every build builds on itself.
Price: Free tier; Pro $20/mo ($17/mo annual). What to charge: The Claude work is baked into the build price — it's your fulfillment tool, not a line item.
Zapier
Zapier handles the integrations your build stack can't reach — legacy apps, niche SaaS, or clients who won't approve a new tool. The Starter plan at $19.99/mo annual handles most SMB workflows, and the ecosystem is wide enough to bridge whatever the client's stack throws at you.
Price: Free (100 tasks/mo); Starter $19.99/mo annual. What to charge: $1,500–$3,000 per automation audit and build, $300–$500/mo maintenance.
A workflow that sells
The intake-to-automation pipeline — the model that closes best for a non-technical consultant:
- Claude maps the process. Run a 60-minute discovery call and load the client's documents into a Claude Project. Use Claude to draft a process map and identify the 3 workflows with the highest friction — that list is the proposal.
- Build in Gumloop or n8n. LLM-heavy builds (enrichment, classification, generation) go in Gumloop; multi-app orchestration or clients who want self-hosting go in n8n. Make fills the gaps.
- Test on real data, deliver with Loom. Record a 10-minute walkthrough of the live workflow running on the client's actual data. That video is your case study — get permission before you post it.
- Retainer = monitoring + iteration. The monthly keeps the workflow maintained as the client's data and prompts drift. Most retainer work runs under 2 hours/month.
Sell the audit and design phase as a $750–$1,500 paid first step; the build follows at $1,500–$5,000 depending on complexity. The paid discovery almost always closes the build.
The money
The standard math: one simple Gumloop build ($1,500) plus one n8n managed engagement ($2,500 + $750/mo) lands $4,000 upfront and $750/mo recurring — off a $70/mo tool stack. Three retainer clients and you've got $2,250/mo in baseline revenue before you touch another project.
The ceiling goes up fast: document processing, multi-step lead pipelines, and AI-agent workflows for regulated industries (legal, finance) bill $5,000–$15,000 for the initial build. You don't need to build any of that alone — see the how to start an AI automation agency playbook, and the broader AI tools for consultants stack if you're packaging other services alongside.
AI Operator Academy is where operators get the build frameworks, the pricing models, and a peer group that's already shipping these engagements at scale — $999/yr.
FAQ
Do I need to know how to code to sell AI automation?
No. Gumloop and Make are fully no-code. n8n has a code node if you need it, but most SMB workflows never touch it. The billable skill is process diagnosis — understanding what's broken and what the right tool is — not writing Python.
What's the difference between Gumloop, n8n, and Make?
Gumloop is AI-native: it's built around chaining LLM steps and shines on workflows that need classification, generation, or enrichment. n8n is the self-hostable infrastructure choice — best when the client wants to own the stack or when you're building complex multi-system agents. Make is the broadest integration layer: 2,000+ apps, the easiest to prototype on, and the right call when the client already runs five SaaS tools that just need to talk to each other.
How do I price a build I've never done before?
Anchor at $1,500 for anything that takes under 8 hours to build, $2,500–$5,000 for anything with multiple moving parts or AI steps. Clients expect software prices, not hourly rates — the $2,500 build that takes you 4 hours is still $2,500. Your confidence in the price signals your confidence in the outcome.
Can I outsource the builds and just sell the audits?
Yes. The "I sell the audit, someone else builds it, I take a margin" model is a legitimate path — several operators inside AI Operator Academy run it this way. Keep the discovery, the proposal, and the client relationship; find a technical operator to fulfill at 20–30% subcontract margin. See what is an AI automation agency for how that model works at scale.
What's a realistic first engagement to pitch?
A lead enrichment workflow: take a prospect list in a spreadsheet, run each company through a scrape-and-summarize Gumloop flow, and output an enriched CSV with a one-paragraph company brief per row. Clients have seen this on LinkedIn and want it; the build runs 3–5 hours; and you charge $1,500–$2,500. Pitch it as a flat-fee "AI prospecting build," deliver in a week, and record a Loom demo before handoff.