Ep 19: How to talk to AI models

Written By Jenna Gaidusek

Listen to the Podcast Episode for a deeper dive

How to Talk to AI Models | AI for Interior Designers™
AI for Interior Designers™ Podcast

How to Talk to AI Models

The single biggest factor in how useful AI is for your practice is how well you communicate with it. Jenna walks through six concrete techniques for getting dramatically better responses — with design-specific examples for each one.

Ep. 19 Solo Episode ✦ Jenna Gaidusek July 2024
This blog was written using AI as a recap from the recording, then edited by the author for accuracy and details.
Key Takeaways
  • Treat AI as a conversation, not a search engine. The designers who get the most value from ChatGPT are the ones who build on responses, ask follow-ups, and iterate — not the ones who ask one question and move on.
  • Specificity is the most important variable in prompt quality. Vague questions produce generic answers. Specific questions — with context, constraints, and a clear goal — produce usable, tailored responses.
  • Context is what separates an AI acting as a general assistant from one acting as a knowledgeable collaborator on your specific business. The more you tell it about who you are and what you are trying to accomplish, the more targeted its output becomes.
  • Breaking complex questions into parts is not a workaround — it is better practice than trying to ask everything at once. Sequential prompts produce more focused and useful answers than compound questions.
  • Iteration is built into the process. The first response is rarely the final answer; it is the starting point for a more productive conversation. Asking the AI to refine, expand, simplify, or reframe its output is the normal workflow, not a fallback for bad results.

What AI Models Actually Are — The Foundation

Understanding what you are working with changes how you interact with it. AI models like ChatGPT are trained on enormous amounts of human-created text — books, articles, websites, transcripts, research papers — which gives them broad knowledge and the ability to generate human-like responses across a huge range of topics.

The practical implication: you can interact with them as you would a very well-read colleague who knows a great deal about many things but knows nothing specific about you, your business, or your clients unless you tell them. The more you tell them, the more the output reflects your specific situation rather than a generic answer to a generic question.

"You can interact with AI models as if they were a knowledgeable colleague — one who needs context about your specific situation to give you the most useful advice."

— Jenna Gaidusek

This is why the communication techniques in this episode matter so much. AI is not a lookup tool that returns fixed answers — it is a generative system that responds to what you give it. Better input produces better output, consistently and predictably.

Six Tips for More Effective AI Communication

01 Be Specific
Detailed, specific questions produce better responses than vague ones. The AI generates based on what you provide — if you give it a broad question, it produces a broad answer. If you give it a precise question with clear parameters, it produces a focused, usable response. Instead of: "How can I improve my design business?" Try: "What are three effective strategies for a small residential interior design firm with one designer to attract clients in the $50,000–$100,000 project budget range?"
02 Provide Context
Background information transforms generic output into tailored advice. Tell the AI who you are, who your clients are, what your practice looks like, and what you are trying to accomplish. This context shapes every part of the response — the tone, the assumptions, the specifics of what it recommends. Context to include: "I am a solo interior designer in Charleston, SC specializing in coastal modern residential projects. My average project budget is $75,000–$150,000 and my clients are primarily second-home owners. My current hourly rate is $175..."
03 Use Clear Language
Avoid dense industry jargon and complex compound questions. Clear, direct language produces clearer responses. If you are asking about something technical, explain what you mean rather than assuming the AI shares your specific industry terminology — which it may interpret differently than you intend. Example: "Write a follow-up email to a client who attended a design presentation last week but has not responded to confirm whether they want to move forward. Tone: warm, professional, not pushy."
04 Break Down Complex Questions
If your question has multiple parts, ask them sequentially rather than all at once. A question with three sub-questions will receive three partial answers; three separate focused questions will each receive one complete answer. The total output from sequential prompts is almost always better than a single complex prompt. Instead of asking everything at once, sequence it: First: "Help me define my ideal client profile for a luxury residential practice." Then: "Based on that profile, what platforms are they most active on?" Then: "What content would resonate with them on those platforms?"
05 Seek Clarification
If a response is unclear, too general, or uses terms you are not familiar with — ask the AI to explain it differently. It is designed to assist and will reframe, simplify, or expand on its output in response to a follow-up. This is a normal part of the workflow, not a sign that something went wrong. Useful follow-up phrases: "Can you explain the third point in simpler terms?" / "What do you mean by [specific term]?" / "Give me a concrete example of how that would work for a residential interior design firm."
06 Engage in a Dialogue — Iterate
The most productive AI interactions are conversations, not single exchanges. Ask follow-up questions, build on responses, redirect when the output is not quite right, and refine toward what you actually need. The first response is a starting point; subsequent exchanges are where the real value is generated. Iteration examples: "That's helpful — now make it more concise." / "Rewrite the second package with a higher-end positioning." / "Give me three alternative versions of that email subject line."

Putting It Together — A Practical Workflow Example

Jenna's worked example in this episode is defining new services for a design business — the same task demonstrated in Ep. 20 with the four-model comparison. Here is how the conversation-based approach plays out step by step, applying all six communication principles.

A Conversation-Based Service Development Workflow
1
Provide context first: your current hourly rate, existing services, typical client type, and geographic market. Do not ask anything yet — just set the foundation.
2
Ask specifically: "Based on what I've told you, suggest three tiered virtual design service packages with deliverables and pricing." One specific question, not a compound request.
3
Evaluate the response and follow up: "The middle tier feels thin — what deliverables would make it feel like a complete service rather than a partial one?"
4
Extend the conversation: "Now help me write a one-paragraph description of each tier that I could use on my website. Tone: professional but warm, not salesy."
5
Move to next phase: "Based on these three packages, suggest a 30-day content calendar I could use to announce and promote the new services on Instagram."

The whole workflow — service design, package descriptions, launch content calendar — can be accomplished in a single conversation thread using these principles. Each step builds on the previous one, and the AI carries the context of everything you have established rather than starting fresh each time.

Frequently Asked Questions
More context is almost always better than less, up to the point where you are providing information that is genuinely not relevant to the task. A good rule of thumb: include anything that would change how a knowledgeable colleague would advise you. Your market, your typical client, your current pricing, your practice size, your specific goal for the output — all of these shape the response in useful ways. What you can leave out: exhaustive background on topics the AI already knows, detailed explanations of standard industry concepts, and context that is interesting but not relevant to the specific question. For regular interactions on the same topics, a custom GPT configured with your standard context saves you from re-entering it every time.
Do not start over — redirect within the conversation. Tell the AI specifically what was wrong with the response and what you need instead: "That's not quite what I was looking for. I need [specific thing], not [what it gave you]. Try again with that in mind." This is more efficient than starting a new conversation because the AI retains the context you have already established. If the response was off-base because your original prompt was unclear, restate the prompt with more specificity before asking it to retry. Genuinely off-base responses are usually a sign that the prompt lacked specific context, not that the AI cannot handle the task.
Within a single conversation thread, yes — ChatGPT maintains context across the entire conversation, which is why the dialogue approach Jenna describes works so well. Across separate conversations, standard ChatGPT does not retain memory unless you have memory features enabled (available in some ChatGPT plans). The practical implication: do your related work within a single conversation thread rather than starting new conversations for each task. If you frequently work on the same types of tasks, a custom GPT configured with your standard context means you do not have to re-establish it each time — it is built into the tool from the start.
The most effective workflow for writing: use AI to produce a first draft, then edit it heavily rather than publishing what the AI generates directly. AI-generated text tends to have predictable patterns — a certain rhythm, certain phrase choices, certain structural conventions — that become recognizable with exposure. Heavily editing the draft produces content that sounds like you rather than like AI. Specific writing applications where this workflow is most efficient: client email drafts, proposal boilerplate sections, social media captions, blog post outlines, project descriptions for your website. Specific applications where AI is less useful: highly personal content that requires your specific voice and perspective, anything requiring information about your specific current projects, and anything where factual accuracy about your specific practice is critical without verification.
Three approaches in order of increasing effectiveness: describe your voice in the prompt ("write in a warm, direct, professional tone — not corporate, not casual, approachable but authoritative"); provide examples of your existing writing and ask it to match the style ("here are three paragraphs from my website — write the new service description in the same voice"); or create a custom GPT configured with your voice guidelines and examples so the style instruction is built into the tool rather than something you have to specify each time. The custom GPT approach is the most efficient for ongoing content production — set it up once and it maintains consistent voice across all outputs without additional prompting. For reference, Ep. 20's service packaging example demonstrated how providing specific business context (rate, services, deliverables) dramatically improved output relevance — the same principle applies to voice.
 
 

Disclaimer: This blog was written using AI as a recap from the recording then edited by the author for accuracy and details.

 
 
Jenna Gaidusek
Previous

Ep 20: Using AI to Fine-Tune Your Design Services and Pricing Strategy
Next

Ep 18: Is AI Coming For Your Design job?