Edward Morris, CEO and Lead Prompt Engineer of Enigmatica.
At the end of my first year building an AI consultancy, I did something most founders avoid: I sat down and read my bank statements line by line, transaction by transaction. Which, arguably, is the worst version of a storybook there is.
Each transaction was a chapter. Each month, a plot twist. Some decisions aged well. Others, less so. But one thing became clear to me: I had built something real in a space where most people were still trying to figure out where to start.
In fact, as my accountant and friend who runs his own firm and was helping with my first fiscal year pointed out to me, I had expected to pay more. I expected traction and growth. But how many people entered AI expecting the same thing and ended the year with nothing to declare?
We have all seen the promises saying you can get into AI without any technical background. No code, no problem. That part is true. The part that is missing is what comes next.
Like many others, I entered AI through words, not code. I was a copywriter. Then those same words became instructions. Prompts. Systems. Over time, that evolved into enterprise work, supporting organizations with implementation, adoption and real-world deployment of AI tools.
The tools may change, but the craft doesn’t. So why does building an AI company work for some and not for others? Here are five lessons I learned that can determine whether an AI career succeeds or falls short.
1. Prompt engineering isn’t dead.
It seems like every few months, someone declares the death of prompt engineering. And every few months, it quietly ignores the obituary.
AI systems still need direction. Whether it is a chatbot, an AI agent or a multi-model workflow, outcomes are shaped by how well the problem is articulated. Better models do not fix unclear thinking. If you hand a chef a vague sentence, you should not be surprised when the dish comes back wrong.
In enterprise environments, this can compound quickly. A vague instruction does not produce just one bad output; it can products hundreds, sometimes thousands. Suddenly, your efficiency tool is creating more work than it removes.
Your interface may feel conversational, and your systems may feel autonomous. But this principle remains: “Clarity in, quality out.” Your overall story still depends on how well you write the scenes.
2. Upskilling alone doesn’t drive adoption.
There is a common assumption that if you train a team on AI, transformation will follow. As if one session turns a department into Tony Stark overnight.
It does not. Many organizations have already invested in tools and training. Employees understand what the tools can do. They have seen the demos. And yet, according to Recon Analytics, only 1 in 4 of the 120,000 Americans surveyed use AI tools daily, with a mere 19% claiming to use it weekly.
I find the issue is usually not capability; it is motivation. Most tools are introduced with possibility: “Look what you can do!” But in my experience, adoption happens when the message becomes about removal: “Look what you no longer have to do!”
Time is the real currency inside a business. Remove grunt work and low-value tasks, and you can expect behavior to change quickly. It is the same principle I learned as a copywriter. An effective AI deployment should delete workload rather than adding to it.
3. The gap between theory and implementation is wider than it looks.
AI content and creators are everywhere. Frameworks, prompt libraries, templates—it can feel like everything you need is already available. But understanding AI and implementing it are very different things.
Early in the AI boom, there was a flood of ready-made solutions, from prompt packs to playbooks to prebuilt systems. Many were well-intentioned, but in my experience, few survive contact with real organizations. This is typically because prompts that work great in one context don’t necessarily work in another. Implementation introduces friction. Data is messy. Permissions are inconsistent. Systems do not align. People resist change. None of that appears in a template.
For leaders, this creates risk. Decisions can end up being shaped by content that has not been tested under pressure. It reads well. It sounds convincing. But it has not been proven. So my advice is this: Don’t adopt a tool, prompt or system that has never been field-tested, no matter what results it promises.
4. AI doesn’t replace thinking.
There is a belief that AI reduces the need for thinking. In practice, I’ve found it does the opposite. AI amplifies whatever you bring into it. Clear thinking produces strong outcomes, while unclear thinking produces nonsense. “Garbage in, garbage out” did not disappear just because the models got better.
I’ve seen this in business environments. Two people use the same tool: One produces something useful, the other produces something unusable. The difference is not the tool; it is the thinking behind it. AI is a great mirror.
5. Accessibility doesn’t mean simplicity.
AI has lowered the barrier to entry. But lowering the barrier does not remove the work required to cross it. It is easier than ever to start, which creates the illusion that mastery should follow quickly. When it does not, people assume something is broken.
In most cases, however, nothing is broken. You still need to understand problems, structure ideas, refine outputs and iterate. The tools have accelerated the process, but they haven’t replaced it. The first chapter is often the easiest to write, but the rest of the book still requires effort.
Final Thoughts
When I looked back at that first year, transaction by transaction, it would have been easy to focus on the outcome. However, the real value I found was in seeing how those results were written: through iteration, mistakes and learning how to think in a way that AI could execute. Not shortcuts.
AI may generate output, but the quality of that output is still authored and recognized by the person behind it. Give a hammer to a carpenter, and you get structure. Give it to a sculptor, and you get art. The tool is the same. The outcome is not. And just like those transactions, line by line, AI reflects the person holding it.
The system writes what you tell it to. So the real question is not what the tool can do; it is whether your expectations are clear enough for it to deliver.
Forbes Business Council is the foremost growth and networking organization for business owners and leaders. Do I qualify?
