Over the first year of my MBA, I found myself as the de-facto AI person in many circles. Classmates were curious about what LLMs could actually do, startup founders wanted help with AI strategy, and friends were navigating career questions around a field that was changing under their feet.
These were not surface-level chats. I was consulting on startup strategy, designing proof-of-concept tools, and acting as a thought partner for people trying to figure out the “so what” of AI.
Eventually, a question began to form:
What if I turned this into something more formal: a consulting agency focused on real-world AI solutions?
The market is real, but fragmented
One surprising lesson was that demand for practical AI solutions is real, especially in small and medium-sized companies. These companies often do not have internal AI talent, but they do have clear operational pain points.
The problem is not always technical feasibility. Often, the problem is search friction: people who need practical AI help and people who can build it rarely find each other at the right moment.
Trust unlocks value
Clients do not just want technology. They want certainty.
Before they give you access to their data, processes, or decision-makers, they want to know:
- that you understand their domain
- that you have done this before
- that they will get a clear return
- that you will not misuse their trust, time, money, or data
This creates a real obstacle. Not because the technology does not work, but because clients often do not feel confident enough to take the leap.
Trust is not a bonus. It is the currency that makes AI consulting work.
Breaking the chicken-and-egg problem
Early AI consulting has a familiar problem: clients want proof of value before they commit, but you need clients in order to build that proof.
There are two practical ways through that loop.
First, start close to your network. Work with people who already trust you, solve real problems, and build small but concrete proof points.
Second, invest in framing. Strategy, storytelling, financial modeling, and domain-specific language are not decorations around the technical work. They are often the thing that lets the technical work become valuable.
Where the real value sits
Through these conversations, I realized something important:
Tech is the delivery mechanism. The value is created around it.
Successful AI work depends on discovery, business understanding, ROI analysis, trust, and the ability to explain why a system should exist in the first place.
Those things cannot be outsourced to a Python script. They require people who know how to listen, learn, and build relationships as much as they know how to work with models.
Why I chose not to scale it full-time
After a year of exploring the space, I decided not to build a full AI consulting agency.
The cold outreach, long sales cycles, market validation loops, and operating machinery required to scale an agency are real work. I respect people who do that work well. I just realized that my motivation was elsewhere.
I found more energy in a simpler pattern: helping people who reach out directly, solving real problems in thoughtful conversations, and building systems when the work aligns with my curiosity and judgment.
Lessons I still carry
For people building in this space, a few lessons have stayed with me:
- Start with real customer problems, not hypothetical ones.
- Expect to spend more time on trust and storytelling than on code.
- Validate demand before you build the product.
- Do not underestimate how much work it takes to sell, even when the technology is strong.
- Most of the value is created around the technology, not inside it.
That lesson now shapes how I think about production AI systems more broadly. The model matters. But so do the workflow, the proof, the framing, and the humans who have to trust the result.