5 Principles for Working With AI Support

One of the most useful effects of artificial intelligence is not that it makes a beginner look experienced.

It is that, when used well, it helps a beginner behave with more structure, more caution and better questions.

I recently experienced this outside a purely technical context. I was preparing documents for a real estate investment involving land acquisition and the possible construction of new properties. It was not a domain where I could claim deep experience.

Before a first meeting with the owner of a construction company, I sent an email with the documents I had collected, the points that seemed important, the issues that needed further analysis and the assumptions behind the possible operation.

Her reaction surprised me. She said it was clear that this was not my first time dealing with that kind of operation.

In reality, it was.

The Humility Behind AI Use

My answer was simple: yes, I am new to this. But now we have artificial intelligence. And I try to remain humble.

Then I added the more important part: You still need to know how to use it.

That distinction matters. AI can make it easier to enter an unfamiliar domain, but it does not remove the need for judgment. It does not turn a first-time investor into a construction expert, a junior developer into a senior engineer, or a manager into a cybersecurity architect.

What it can do is help you organize uncertainty. It can help you discover the questions an expert might expect you to ask. It can make visible the documents, risks, constraints and scenarios that would otherwise remain implicit.

Used in this way, AI is not a shortcut around competence. It is a training environment for competence.

Do Not Start With the Task

In many AI training sessions, I see a recurring pattern. A person receives an instruction such as: prepare this document, summarize this information, create this plan, analyze this case. Then they paste the instruction into an AI tool and turn it into a prompt.

That can work for simple execution.

But it is often not the best way to learn or decide.

A better starting point is to ask the AI what it needs to know in order to help. This changes the relationship immediately. The AI is no longer treated only as an executor. It becomes a collaborator that helps define the problem space.

For example:

• What do you need to know before you can help me evaluate this?
• Which documents or assumptions are missing?
• What are the main risks I should investigate?
• Which questions would an expert ask at this stage?
• What alternative options should I compare before deciding?

These questions are not sophisticated because they are long. They are sophisticated because they invite the system to help structure the work before producing the output.

AI Should Open the Problem Space

In the real estate case, the useful questions were not only about which documents were needed for a purchase. They were broader.

What could go wrong? What should be checked before committing? What land-use options might make sense after construction? What would be the difference between selling, renting, short-term tourism, or another destination? How would each option change the business plan?

Once the questions became broader, the AI was useful in a different way. It helped create comparison tables. It helped separate assumptions from facts. It helped identify points to validate with professionals. It helped me prepare a business plan around payback time, possible use cases and trade-offs.

The value was not that the AI provided the final answer. The value was that it made the conversation with an expert more productive.

That is a strong pattern for organizational adoption. AI should not replace expert validation when the stakes are real. It should help people arrive at expert conversations with better context, better structure and better questions.

The Difference Between Prompting and Capability Building

This is why organizations should be careful when they reduce AI adoption to prompt writing.

Prompting is useful, but the deeper capability is learning how to think with the system.

If people ask only for the task, they receive the task. If they ask for the risks, the missing information, the possible scenarios, the evaluation criteria and the expert questions, they receive a much richer working environment.

This matters for managers, analysts, developers, executives and operational teams. In every unfamiliar or high-complexity domain, the first output is rarely the most important output. The most important output is often a better map of the problem.

That map helps people decide what to verify, who to involve, what to compare and where risk may be hiding.

A Practical Pattern for AI-Assisted Work

A simple workflow can make AI use more valuable, especially when someone is working in a new domain.

1. Declare your level of knowledge.
   Tell the AI what you know, what you do not know and what outcome you are trying to reach. This reduces false confidence and makes the collaboration more honest.
2. Ask what context is needed.
   Before asking for the final task, ask which documents, constraints, definitions, stakeholders, deadlines or assumptions should be clarified.
3. Ask for risks and blind spots.
   Invite the system to identify what could go wrong, what an expert would verify and what should not be assumed.
4. Compare options explicitly.
   Use tables and decision criteria to compare alternatives. Do not ask only whether one preferred option is good.
5. Validate with domain experts.
   Treat the AI output as preparation for better expert conversations, not as a substitute for professional judgment.

Why This Matters for Organizations

The same principle applies inside companies adopting advanced technologies.

If AI is introduced only as a tool for faster output, teams may learn to automate fragments of work without improving the way they think, coordinate or evaluate risk. They may produce more material, but not necessarily better decisions.

If AI is introduced as a collaborator for structured thinking, teams can build internal capability. They learn how to surface assumptions, ask for missing context, compare scenarios, prepare expert conversations and connect outputs to governance.

That is the bridge between tool use and organizational adoption.

The goal is not to make people less humble. It is to make them more capable while they remain aware of what they still need to learn.

Read more ideas and field reflections on AI, advanced technologies and internal capability.