The line-item that made me walk away from the proposal
A pest-control company I worked with got a proposal from a "boutique automation consultancy" for a lead-triage system.
The scope was 22 pages. The price was $28,000. And buried on page 14 was a line that said "Project Manager: 40 hours @ $180/hr."
The actual automation—the thing that routes inbound leads, scores urgency, and drops them into the tech's calendar—would take maybe 12 hours to build in n8n.
But they were billing $7,200 for someone to forward emails between the builder and the client.
That's when I realised most custom AI automation agencies aren't selling automation. They're selling coordination overhead.
What you're actually paying for when you hire an automation agency
Let's break down a typical $25k engagement with a multi-person shop:
- Discovery & scoping: 8–12 hours of meetings. Usually led by someone who won't touch the build.
- Project management: Weekly check-ins, status decks, change-request triage. Billed at $140–$200/hr.
- Design & documentation: Flowcharts, wireframes, handoff specs. Often created for internal handoff, not for you.
- Development: The actual build. Usually 20–40 hours of hands-on work.
- QA & revision cycles: Another 8–12 hours, stretched over weeks because the builder is juggling six other projects.
Add it up and you're paying $15k–$18k for coordination, $7k–$10k for the work.
The math flips when you work with a one-operator shop. No project manager. No internal handoff. You talk to the person who writes the workflow, and they ship it in 2–3 weeks because there's no coordination tax.
Why "custom AI automation agency" usually means "we'll build anything" (and that's the problem)
Most agencies pitch themselves as platform-agnostic. They'll build in Zapier, Make, n8n, Python, whatever you want.
Sounds flexible. In practice it means:
- Every project starts from scratch. No library of tested components.
- The builder assigned to you might be learning the platform during your build.
- You get a Frankenstein system that works on delivery day, then breaks the first time an API changes.
A good custom AI automation agency should have opinions. Not because they're stubborn—because they've seen what breaks at scale.
For example: we only build in n8n. Not because it's trendy, but because self-hosting means no surprise monthly bill when you scale from 100 to 10,000 runs. And because JavaScript expressions let us handle edge cases without needing Zapier's $600/month webhook plan.
If an agency will build in "whatever you prefer," ask them how many production systems they've shipped in that tool. If the answer is less than ten, you're paying them to learn.
The three questions that separate a builder from a sales team
When you're evaluating a custom AI automation agency, ask:
-
"Who will I talk to during the build—and is that the same person writing the code?"
If they say "your dedicated project manager will coordinate," that's a red flag. You want direct access to the builder. -
"Can I see a workflow you shipped last month?"
Not a case study. An actual exported workflow or a Loom of the system running. If they can't show you real work, they're selling strategy decks. -
"What happens if I need a change six weeks after launch?"
Good answer: "I charge $X/hour for revisions, typical fix is 1–2 hours."
Bad answer: "We offer a retainer starting at $2k/month for ongoing support."
The third question is the nuclear one. A lot of agencies want you locked into a retainer because that's where their margin lives.
A well-built system shouldn't need a retainer. It should need occasional tweaks when your CRM changes or you add a new product line.
Why one-operator shops ship faster than ten-person teams
Here's the calendar reality of a mid-sized agency build:
- Week 1: Kickoff meeting. PM takes notes, sends summary.
- Week 2: PM briefs builder. Builder has questions. PM schedules follow-up with client.
- Week 3: Follow-up call. Builder starts work.
- Week 4: Builder finishes v1, hands to QA. QA finds issues, sends back.
- Week 5: Builder revises. QA approves. PM schedules demo.
- Week 6: Demo. Client requests two changes. Repeat.
Total build time: maybe 30 hours. Total calendar time: 6–8 weeks.
Now here's a one-operator build:
- Day 1: 90-minute scoping call. I ask questions, you answer, I sketch the workflow live in Excalidraw.
- Days 2–10: I build it. You get a Loom every 2–3 days showing progress.
- Days 11–14: You test. I fix. We go live.
Same 30 hours of work. Three weeks start to finish.
The difference isn't skill. It's communication hops. Every handoff adds three days of latency.
When you work with a single operator who ships every build personally, there are zero handoffs.
What "custom" should actually mean
Most agencies define "custom" as "we'll build whatever you describe in the kickoff."
That's not custom. That's bespoke—and bespoke usually means expensive and fragile.
Real custom work means:
- We dig into your actual process before we design anything. Not what you think should be automated—what actually eats 12 hours of someone's week.
- We tell you when a workflow is a bad idea. If your "automation" requires 14 conditional branches and three manual approvals, we'll say so. Better to simplify the process than automate the mess.
- We build for the next twelve months, not just launch day. That means logging, error handling, and a way for you to see what's running without opening the workflow editor.
If you want to see which of your workflows are worth automating in the first place, run them through the Repetitive Task Cost Calculator. It shows annual cost of manual work vs. build cost. Most teams are shocked to find they're spending $18k/year on a process that could be automated in 15 hours.
The agency model that actually works for small teams
Here's the structure that makes sense if you're a 5–25 person business:
- Fixed-scope build: One system, one price, shipped in 2–3 weeks. No retainer required.
- Direct access to the builder: You ask questions in Slack or Loom, you get answers in hours, not days.
- Handoff document + training: At the end you get a Loom walkthrough and a one-page doc that explains how to tweak the workflow when your process changes.
That's it. No ongoing contract unless you want more systems built.
Most custom AI automation builds we ship are in the $4k–$9k range. That's not because we're cheaper—it's because we're not billing for project managers, QA testers, and revision cycles inflated by internal miscommunication.
When you should pay more (and when you shouldn't)
You should pay $15k–$30k if:
- You're automating something mission-critical with compliance requirements (healthcare, finance, legal).
- You need a custom-trained model, not just API orchestration.
- You're integrating with a legacy ERP that requires reverse-engineering SOAP endpoints.
You shouldn't pay $15k–$30k if:
- Your automation connects standard SaaS tools (CRM, email, Slack, Google Sheets).
- The workflow has fewer than eight steps.
- You're a team of less than twenty people.
If an agency quotes you $20k for a lead-triage system that uses Webhooks, OpenAI, and your CRM API, they're overcharging. That's a $5k–$7k build, max.
Run your site through the Automation Opportunity Scanner and you'll get three ranked automation ideas with ROI math. It takes 60 seconds and shows you what's worth building before you talk to any agency.
One operator, no coordination tax, shipped in three weeks
Sinqra is a one-person shop. I scope, build, test, and ship every system.
No sales team. No project manager. No QA department. Just me, n8n, and a Slack channel where you can ask questions whenever you want.
If you're tired of agencies that bill $25k and deliver eight weeks later, let's talk. Most builds are 2–3 weeks, fixed scope, and you'll talk to the same person from kickoff to launch.
Book a 30-minute scoping call and we'll figure out what's worth automating, what it'll cost, and how fast I can ship it.