EPISODE 010
$50,000/month Helping Small Businesses Implement AI
Builder profile
In episode 010, Simon Bergeron explains how he built Lemonbrand into roughly a $50,000 a month business as an AI Sherpa for small businesses. He comes in, finds where AI can help, builds it, trains the team, and keeps it running. Simon has worked with 110 companies across more than 60 industries, and he walks through his first client, the three projects clients ask for most, how he prices his work, his tech stack, and why he thinks selling software is a dying profession.
What Lemonbrand does
Simon calls himself an AI Sherpa. A Sherpa knows the pitfalls on the mountain and the spots where the rocks are crumbly, and guides you over. In practice he acts as the AI department for small businesses. He asks where they want to go, builds agents to take over some workflows, and trains their team to use AI effectively. He learned by spending 10 to 12 hours a day, on average, for the last four years in Claude, Codex and open source models. His view is that you learn by doing.
The first client
Simon was an employee when he asked his company if he could do this full time. They converted his contract into a one-year fixed monthly retainer. That cushion gave him room to find clients two, three and four, which were all local businesses. After that he expanded into outbound to find people farther away who paid more.
How he finds clients
For a while he did outbound himself, sending thousands of emails a week and doing LinkedIn outreach. Now he audits businesses and maps their workflows and the low hanging fruit. Right now, 80% of his new clients come through white label partners who use his audit map or his fulfillment strategy, while he supports them behind the scenes. White label means someone sells your service under their own brand, like a supermarket selling another company's chips under its own packaging.
The team behind it
Simon is solo operationally. The name Lemonbrand comes from his name and his spouse's name, Leah. He has four freelancers on the payroll and may add more, but he wants to stay small. The business is bootstrapped, and he is not looking for outside funding because he wants to keep control of the vision.
The three projects clients ask for most
First is new business: answering leads the moment they come in with a bot, for example asking a hot tub mover's prospects to send pictures of the site. Second is data entry, moving numbers from one place into a spreadsheet. Third is reporting, pulling numbers from different apps into one place so leaders can make informed decisions.
Which industries it works for
He has worked across 110 companies and over 60 industries, and no single industry pulls ahead. He compares AI to the internet and email: a base layer you shape to each industry. Marketing and professional services are moving fastest, because AI is good at bureaucracy and documents. Healthcare and legal lag behind because they are highly regulated and a person has to sign off. Local service businesses mostly need better organization and follow up. Many have gold in their inbox from people who contacted them in the past and never heard back. Digital businesses mostly want help with social media, such as posting 10 times a day.
Making a first client more money than you cost
Simon's advice for a first client is to make them more money than they spend on you within the first 30 days. His first client was a gym. Most gyms have past clients they never contact, so he set up a system to text them. You can do it manually at first and then automate it. If you bring in more recurring revenue that month than you cost, it is easy to expand into more work and ask for more money.
How he prices AI work
He starts from first principles with what it costs to deliver. If your tools come to $500 a month, your first client should cover that, plus what you pay yourself, plus a small margin for negotiation. His example is tools at $500, paying himself no more than $500 on the first client, and quoting $1,500. His current tiers are a $500 a month maintenance tier, a first, one-time piece of work at about $1,500 for a small team and $2,500 or $5,000 for a larger team, and retainers at $2,500, $5,000 and $10,000 a month. The $1,000 tier is training only. At $2,500 he adds agents, at $5,000 he adds database management such as a knowledge base or second brain, and at $10,000 he runs multi-day sprints with thousands of agents.
The tech stack
To start, he suggests one AI tool of choice on the $100 tier of Cursor, Claude Code or Codex (ChatGPT), upgrading when you hit weekly limits. Get a Gmail account to keep your professional life separate from your personal one. He hosts on Vercel and uses GitHub. For basic automations he suggests n8n, which has a good MCP tool that lets your AI build the automations for you.
What the business earns and spends
Simon shares his numbers publicly, as approximations. In the month we recorded, he made about $52,000, spent about $24,000 on costs, paid himself about $10,000, and reinvested the rest. His biggest client pays around $10,000 a month and his smallest around $500. Most of the costs go to inference: model companies and APIs used to deliver the work. He has spent $0 on ads so far because he is not sure he could handle the demand. He has set aside $40,000 to start on ads, probably next year. He is also shifting from steady monthly retainers to an upfront payment for the audit map followed by a retainer, so he can pay white label partners.
Training clients and owning the agent
Every client has asked him to train their team, and he does it freely as part of the value. He does not think agents should be sold as a subscription. He compares it to hiring an employee that someone else owns and you never see, a black box. He thinks companies selling agents that way will eventually go under, because it keeps clients hostage to the value they generate. He prefers to teach people so they feel ownership and become agent managers who delegate tasks.
Why selling software is dying
Simon says selling software is a dying profession, within five years at most, and he bets on two. He points to clones of Photoshop and After Effects, and says you can reverse engineer software with agents. The hard part is change management: getting people in a company to feel good about AI, feel the productivity gains, and own them. He says it comes down to humans, and the seller's job is to manage expectations and growth.
What to build with AI
Simon does not suggest building a new tool. He suggests finding APIs and connectors that already exist and making them accessible to other people. His example is social media posting tools like Blotato, which let a small business schedule posts and see analytics for roughly $10 to $30 a month. By selling the implementation, he says you can make a 90% margin. He would not start by building a SaaS.
Find Simon
- Lemonbrand: lemonbrand.io
- YouTube: youtube.com/@slbergeron
- LinkedIn: linkedin.com/in/slbergeron
- Substack: slbergeron.substack.com
- Instagram: instagram.com/lemonbrand.io
- TikTok: tiktok.com/@simonbergeron
- X: x.com/SLBergeron
Simon also runs a weekly webinar every Thursday at 1 p.m. Eastern.
