I made $75K selling AI automations to clients. Here's what I'd change if I started over.

I wasn't planning to build an AI automation business. I was freelancing, doing GTM work for SaaS founders, and one of them asked if I could "set up some AI thing" to handle their lead follow-ups. They were losing prospects because their two-person sales team couldn't reply fast enough. ​ I quoted $2,500. Built it over a weekend using Zapier and GPT. Took their average first-response time from 14 hours to under 3 minutes. The client told a friend. The friend called me. ​ That was roughly a year ago. I've done $75K in revenue since then, mostly from small and mid-size businesses that knew AI could help them but had no idea where to start. I want to share what actually happened, because most "I made X with AI" posts skip the parts where it got ugly. ​ My first five clients all came from referrals. I didn't have a website, a portfolio, or a pricing page. Just a WhatsApp message that said something like "the guy who fixed our lead flow." I charged between $1,500 and $3,000 per project. Felt like a lot at the time. ​ It wasn't. I was building custom workflows, integrating three or four tools, handling revisions, jumping on calls, and basically being on retainer for free because I hadn't scoped the engagement properly. One project I quoted at $2,000 ate six weeks of back and forth. That client is the reason I now send a scope document before I touch anything. ​ The first real lesson hit around client number seven. A dentist's office wanted me to automate appointment reminders and no-show follow-ups. Straightforward. But then they asked me to also build a chatbot for their website, connect it to their booking system, and "maybe do something with reviews." The project went from a clean $3,000 build to a mess I was still patching two months later. ​ I stopped saying yes to everything after that. Picked three services I could deliver well and fast: lead response automation, CRM workflow builds, and AI-assisted customer support setups. Nothing else. If someone wanted a chatbot on their site talking to their inventory system, I'd refer them out. Saying no to money felt wrong until I realized the money I was saying no to was always the money that cost me more than it paid. ​ Pricing is where most people in this space leave cash on the table. I know because I did it for months. ​ My early projects were flat-fee. $2,500 to build, hand over, done. The problem is that a lead-routing automation I built in 12 hours was worth the same to me as one I built in 40. The client paying $2,500 for the 12-hour build was getting a steal. The one paying $2,500 for the 40-hour build was getting my full attention and I was getting minimum wage. ​ What fixed it: I started charging a build fee plus a monthly retainer. Build fee covers the setup, usually $3,000 to $7,000 depending on complexity. Retainer covers monitoring, tweaks, and the fact that these systems break quietly. APIs update. Rate limits change. A form field gets renamed and suddenly nothing flows. $500 to $1,500 a month, depending on how many automations the client runs. That retainer revenue is what turned project work into a business with actual recurring income. ​ Right now about 60% of my revenue comes from retainers. The other 40% is new builds. The retainer clients barely contact me most months, but they pay because the one time something breaks at 2am and leads stop flowing, I'm the person who fixes it before they wake up. That peace of mind is worth more to them than the dollar amount on the invoice. ​ The $75K breaks down roughly like this: around 18 clients total, average project size just under $4,200, and eight of those clients are on monthly retainers. Three of the retainer clients have been with me for over eight months. Two of them have referred me to other businesses, which brought in another $11K I didn't have to sell for. ​ Some things I'd tell anyone starting this now. ​ Don't sell AI. Sell the outcome. Not once has a client asked me what model I'm using or whether I'm running n8n or Make or Zapier. They want to know how fast their leads get a reply, how many hours their team saves per week, and whether their follow-up rate goes up. If I pitched "I'll build you a GPT-powered multi-step automation workflow," their eyes would glaze over. "Your leads will get a personalized reply within 90 seconds, 24 hours a day" is what closes. ​ Pick boring businesses. My best clients aren't tech companies. They're dental offices, HVAC companies, real estate teams, insurance brokers. Businesses drowning in manual follow-ups and appointment scheduling with zero internal tech talent. They don't comparison shop. They don't ask for a technical proposal with an architecture diagram. They just want the problem gone. ​ Scope everything in writing before you start. I said this already but it matters enough to repeat. The project that nearly burned me out could have been avoided with a one-page document listing exactly what I'd build, what I wouldn't build, and what counts as a revision versus a new project. I use a dead simple template now. Hasn't failed me yet. ​ Don't build from scratch when a platform handles 80% of it. My early instinct was to code everything custom because it felt more "real." Waste of time. Most client problems are solved with Zapier or Make connected to their CRM, plus a GPT layer for the parts that need language. Save custom builds for the rare client whose problem actually demands it. The other 90% want it done Tuesday, not done perfectly. ​ Charge for maintenance. This is where the business gets stable. A one-time build makes you a freelancer. A build plus retainer makes you a partner they budget for every month. Once a client is on retainer, the switching cost is high enough that they stay. Not because you're trapping them, but because replacing the person who knows where all the wires connect is more painful than the monthly fee. ​ I'm not pretending $75K is life-changing money. Spread across a year, after tool costs and taxes, it's a solid income but not a windfall. What changed things for me is that the pipeline is warmer now than when I started, retainer revenue covers my base expenses before I sell anything new, and the last three clients came inbound without me spending a dollar on marketing. ​ If you're selling AI automations or thinking about starting, I'm curious what's working for you. Especially around pricing. I still feel like I'm figuring that part out. ​