How to Start an AI Consulting Business From Scratch

TL;DR

You can start an AI consulting business by combining existing business skills with AI, identifying a company’s friction, and matching it to an existing feature, connected tools, or custom orchestration. Brandon Gadoci reports growing his company from zero to $80,000 a month within five months while working from home with a couple of contractors and little overhead. Read on for his three solution levels, required skills, and practical approach to client problems.

Transcript

So, I made a new friend on Twitter. His name is Brandon. Why is Brandon awesome? Because five months ago, he started an AI consulting business. >> All my normal skill sets plus AI equal superpowers >> using a lot of the same tools that I talk about on this podcast. And within 5 months, he went from zero to $80,000 a month. Put a number on a proposa... Read More

Key Insights

• An AI consulting business can reach $80,000 a month in roughly five months as a solo operator. Brandon Gadoci did it from home in North Dallas with a couple of contractors and a tiny bit of overhead, and says revenue is still growing every month.
• The qualifying skill set for AI consulting is not a computer science degree. Brandon describes his skills as a mile wide and a few inches deep, taught himself to code without being a former developer, and says all his normal skill sets plus AI equaled superpowers.
• AI is a probabilistic technology, meaning the same question returns different answers each time, directionally accurate but not deterministic. Deterministic code is the if this then that logic software has traditionally used, where one plus one always equals two.
• The most successful agentic AI implementations combine deterministic code with probabilistic AI in a back and forth. Brandon says the agentic AI promise being marketed today is alluring but the technology is not there yet, so this hybrid is where the value currently sits.
• A level one AI solution is showing a client that a feature already inside ChatGPT solves their stated problem. Brandon frames the payoff concretely: the client saves around 30% of their time and gets to spend it on what they care about.
• A level two solution connects existing tools such as Zapier, Make, n8n, or Relevance AI to move information between systems with AI somewhere in the loop. A level three solution requires building and deploying custom orchestration for a complex task.
• Much of what companies call AI is actually just automation. Brandon's filter is to find where AI does something automation cannot do on its own, specifically where a context window lets the system understand nuance rather than follow fixed rules.
• Difficulty is not the same signal as wrongness in business. Brandon's rule is that when you are headed the right way the work will be hard but not complicated, and this venture felt like wind in the sails rather than pushing a boulder uphill.

Questions & Answers

Q: How do you start an AI consulting business from scratch?

Start by combining your existing skills with practical knowledge of AI and focusing on meaningful business results. Brandon Gadoci’s approach is to understand a company’s purpose and friction, then recommend an existing AI feature, connect available tools, or build custom orchestration depending on the problem.

Q: How much did Brandon Gadoci’s AI consulting business grow in five months?

Brandon went from zero to $80,000 a month within five months. He said the business was still growing every month and operated from his home with a couple of contractors and a tiny amount of overhead.

Q: What skills do you need to become an AI consultant?

Brandon says a computer science degree or previous career as a developer is not required. He describes his own skills as broad rather than deeply specialized, taught himself to code, and found that combining his normal skill sets with AI gave him a strong advantage.

Q: What are the three levels of AI consulting solutions?

Level one uses an existing feature, such as something already available in ChatGPT, to solve the client’s problem and potentially save around 30% of their time. Level two connects tools such as Zapier, Make, n8n, or Relevance AI, while level three involves building and deploying custom orchestration for a complex task.

Q: How should an AI consultant identify the right solution for a client?

Begin by asking what the company exists to do and where friction is preventing that work. Then match the problem to the simplest suitable level: an existing AI feature, a combination of available tools, or a custom-built solution.

Q: What is the difference between deterministic code and probabilistic AI?

Deterministic code follows fixed logic, such as “if this, then that,” and produces a consistent result. Probabilistic AI can return differently worded or structured answers to the same question while remaining directionally similar.

Q: What works better than fully autonomous agentic AI today?

Brandon says the strongest implementations combine deterministic code with probabilistic AI and pass work back and forth between them. Deterministic components retrieve or move specific data, while AI handles steps that require nuance or understanding.

Q: How did Brandon know AI consulting was the right business to pursue?

He says the work felt hard but not complicated, which he uses as a signal that he is moving in the right direction. Unlike earlier ventures that felt like pushing a boulder uphill, this business felt as though it had wind in its sails because his skills, curiosity, available time, and market need aligned.

Summary & Key Takeaways

• Brandon Gadoci owns an AI consulting and services company started four or five months before the interview. He is married 23 years, has two kids aged 16 and 15, and lives in North Dallas. He went from zero to $80,000 a month and says he is still growing every month, running the business from home with a couple of contractors and a tiny bit of overhead.

• Before this business, Brandon spent the last three years of his previous job doing what he calls AI operations, a term he credits to Rachel Woods out of Austin. That work was about experimenting with how to bring the technology that arrived in November of 2022 into the enterprise in a way that produces real, meaningful results rather than novelty.

• Brandon sorts AI solutions into three levels. Level one is showing a client that an existing ChatGPT feature already solves their problem, saving perhaps 30% of their time. Level two is cobbling together existing tools such as Zapier, Make, n8n, or Relevance AI to move information around. Level three is building and deploying custom orchestration for a complex task.

• The most successful agentic AI implementations Brandon sees combine deterministic code with probabilistic AI, handing work back and forth. He argues the industry promise of agentic AI is alluring but not yet fulfilled by the technology, so the practical value today comes from that hybrid rather than from fully autonomous agents.