You do not need to be a programmer to use AI. You can start the same way I did: with something you want to figure out.
I am rebuilding Valhalla de Tortuga, our ferrocement sailing boat in Mexico. There are manuals to find, equipment to compare, photographs to sort, and decisions that affect the next stage of the refit. I use AI to help carry that work. But some of the most useful work begins before I give the crew a specific assignment.
It begins with a conversation.
Sometimes I know the destination but have not worked out the route. I talk through an idea. The assistant asks a useful question. I explain what matters, correct a misunderstanding, or change my mind. Together, we turn an idea into a clearer goal.
That conversation is part of the work.
Make It Comfortable
You may prefer a masculine voice, a feminine voice, an androgynous character, or a neutral assistant. You might want warmth and humor, or brief, direct answers. Those are preferences you can express. Where your app offers different voices, the audio voice may need to be selected separately.
I have given our crew names, roles, personalities, and working preferences. I call that programming their personalities. More precisely, I am giving instructions that shape the conversation; I am not rewriting the underlying model.
Comfort matters. It makes it easier to say, “I don't understand,” or “Explain that again.”
A Prompt Is an Instruction
“Prompt” sounds technical. It can be as ordinary as this:
I am new to this. Help me understand this paragraph from my manual. Use plain English, explain unfamiliar words, and ask one question at a time if you need more information.
You do not need a perfect sentence. Explain what you want to accomplish, provide the relevant information, and say what would make the answer useful. Correct the assistant when it misunderstands.
Friendly Does Not Mean Correct
This is one of the most important lessons I have learned: a confident answer can still be wrong.
AI can invent details, misunderstand a photograph, or agree too readily with a mistaken assumption. An invented or unsupported answer is often called a hallucination. It does not become reliable because it sounds polished.
One of my working instructions is:
Accuracy comes before speed. Do not agree just to please me. Check the original sources, separate facts from guesses, challenge my assumptions when necessary, and tell me what you could not verify.
Where an app supports saved instructions or memory, ask it to retain this preference and confirm it was actually saved using the app’s available controls. Do not assume every new conversation, account, or assistant receives it. Our startup handoff reinforces it by explicitly reading the standing orders.
Those instructions help guide the work. They do not eliminate mistakes. I still want evidence—especially when the answer could affect our safety or cost us money.
Give the Crew a Reference Shelf
We organize our manuals, drawings, photographs, and decisions into a reference library. A cloud service can help, provided your assistant has a supported, authorized way to access it.
Our rule is to consult relevant project records first. If the answer is missing, research outside sources and retain a permitted copy or an indexed link with notes for later use. An index tells us where information may be found; it does not mean the document has already been read.
For changing information, such as prices or advisories, an old saved answer needs a fresh check.
Keep the Captain at the Helm
AI can help choose research steps within an assigned job, organize project phases, and explain what remains unresolved. It does not automatically work between conversations, and it should not report something as completed without evidence.
In our mast investigation, we are sharing actual fitting photographs so knowledgeable owners can compare them. We are still asking who made the mast. We have not turned a promising lead into a proven answer.
That is the kind of AI use I want to show: useful, approachable, and honest.
You can begin with one small question, check the answer, and build from there.