Stop searching. Start conversing: how to build real AI skill
Most people approach artificial intelligence like a Google search. They ask a question, skim an answer, then decide the tool is average. The interface invites that behaviour. Real skill with AI comes from a different habit: conversation. Treat AI like a partner you practice with, not a one question search.
Develop your AI Conversation skills
When we search with Google we typically phrase our question with the answer in mind. AI though works differently. It’s driven by conversation which invites nuance. When you sit down prepared for a backand forth dialogue with AI, you get a much more meaningful response. Especially when you add context, examples, constraints, and a clearly defined audience. You can also correct misfires quickly. A conversational approach changes the quality of output and the quality of your thinking. It is less about getting a perfect answer first time, more about refining the ask until it fits the job.
Five prompts a day
Daily practice will improve your conversation skills. Set a small daily target you can hit consistently. Five prompts a day is enough to improve your skills without becoming a chore. Rotate through tasks you already do: responding to a tricky email, following up a meeting with a meaningful summary, or draft a proposal. Keep what works, critique what does not, and iterate. After a month you will have dozens of reusable patterns and a better eye for how you can use AI to help you get work done.
Set your baseline with custom instructions
Most AI tools let you set persistent preferences and context. Define your region, spelling, preferred tone, and the structure you want replies to follow. For example, ask for a one-sentence TL;DR, followed by a short executive summary, then the detailed response. You can also tell AI tools how to challenge you, instructing that you don’t want it just to agree with you. Custom instructions remove repeated friction and lift the baseline response quality for every task.
Learn in the context that matters to you
Skills stick when they serve a purpose that matters. In Aotearoa New Zealand, Te Wiki o te Reo Māori runs from 14 to 20 September 2025. It is a good moment to practise te reo Māori in everyday interactions. Custom GPTs like Kaiāwhina Reo and Whakawhiti Reo from kahu.code provide bilingual support for learners and are built by a Māori father-daughter team. Try drafting a greeting, then ask the tool to correct and explain. Reply to a comment in te reo, then translate it back to check you understood. The goal is learning, not outsourcing.
Language also carries regional nuance. New Zealand flax is commonly called harakeke, while kōrari is used in particular regions. That difference can hint at place and context. Conversation surfaces details and nuance like this that is important both in personal conversations and in conversations with AI.
Words matter: must, should, may
If you work with standards, policy, or risk, tiny words carry weight. In New Zealand standards, “shall” or “must” signals a requirement for conformity. “Should” signals a recommendation. “May” signals permission. When you write prompts, be equally precise. If you want a mandatory rule, say must. If you want optional advice, say should. That precision reduces rework and surprises.
Voice mode and camera inputs can unlock a different way to think with a tool. Talk through a problem and ask AI to summarise what it heard. Point your camera at a whiteboard sketch, a screen, or a photo of a plant you are trying to identify, and ask for a structured description. OpenAI’s voice and image capabilities have been available for some time, and similar features are landing across enterprise tools like Copilot. Use them when typing slows you down.
Sometimes a tool speaks with confidence about features that do not exist. Watch for hedging words like “should see” where “you will see” is expected. If something seems off, say so, and ask for the authoritative source. For product interfaces, that usually means up-to-date docs. For policy, it means the primary standard or Act. For subscriptions, it means the actual cancellation path, not a vague help article.
Choosing the right tool
Different tools are good at different jobs. Treat them like a tools in a workshop. A hammer drives nails, a drill makes holes, pliers grip. ChatGPT, Microsoft Copilot, Claude, Gemini, and others each have strengths. Switch as needed, and go deep on a few you use every day. If someone tells you their favourite is the only correct choice, that is marketing, not an evaluation method.
Have better conversations with AI
State your goal. What outcome do you need, for which audience, by when.
Add constraints. Word count, tone, region, spelling, file type.
Give an example. Paste a sample of “what good looks like.”
Invite questions. Ask the model to ask you three clarifying questions before it writes.
Iterate in public. Keep edits in the thread so you can see how the work improved.
Archive your winners. Save prompts, examples, and outputs to reuse next time.
Focus on people and practice
Practice connects minds. Research shows speaker and listener brain activity can sync during successful communication. No one is suggesting your laptop has a brain, yet conversation builds shared structure between you and the tool. As you practice having conversations with AI, make sure you seek out opportunities to have conversations with your colleagues, friends and family about how you’re learning to use AI.
Treat AI as a conversation partner, practise with purpose, write with precision, and your results will improve.
