Your Best Employee Has Not Been Hired Yet

It was a Friday night, and I had just pulled up to the University of Chicago to pick up my friend Batu for a drink. He had just wrapped up teaching a 3 hour class on AI and I could tell he was mentally beat. In his own words, “students are turning in papers that are often easy to identify as AI-written, making the same obvious mistakes over and over again”. I launched into my own frustrations, the hallucinations, the constant close review of outputs, the times I had trusted AI with something real and walked away disappointed. Then he stopped me cold.

"You can trust AI the same way you trust another person you're assigning work to."

That one sentence sat in the air between us for a moment. My gut immediately wanted to fight it. I started reaching for the holes. One by one, I couldn’t find them.

In the end, the AI producing mediocre results simply hadn’t been trained well enough. It was doing the kind of work you would expect from a poorly trained new hire.

😤 The Way Most of Us Use AI Is Like Hiring Someone and Never Training Them

Here is an honest question: would you hand a brand new hire a complex client deliverable on their first day? No briefing, no context, no sense of your standards, and then judge their entire career on what came back?

Of course not. That would be unfair and frankly a little absurd.

But that is exactly what most of us do with AI.

We open a chat window, type a vague request, get a generic result, and conclude it’s magical but not ready for real work. The problem was never the AI. The problem was the onboarding. Interior design studios, architecture firms, and other businesses are leaving serious productivity on the table. Not because AI is failing them, but because nobody groomed the AI for the job properly.

💡 Meet the Custom AI Agents - Your New Team Member

Every major AI platform gives you the ability to build a custom agent, and that is the secret sauce. Think of it as a version of AI that has been groomed specifically for your business, your voice, and your way of doing things. Most importantly, it understands the specific job you are about to give it.

And here’s the good news. You do not need to be an expert to set this up. It is not rocket science. Most people can build one themselves with a little thought and care.

Here is what each AI platform calls these little worker bees:

• ChatGPT → Custom GPTs

• Claude → Projects

• Gemini → Gems

• Copilot → Copilot GPT

Everyone works with the same trained, configured, and briefed version of AI. Not a blank slate where results change depending on who is typing and the AI system is left guessing how to fill in the gaps.

That is the shift. You are not using AI. You are employing it.

🧩 The 7 Things Every AI Agent (Employee) Needs

Building a custom agent is the onboarding process. The more clearly you set expectations upfront, the better and more consistent the output. Here are seven pieces that you need to put together as your custom instructional prompt for the AI agent.

Examples below are kept short on purpose to give you a starting point. Feel free to get specific within each seciton.

1. Persona and Context - The "Who" and "Where" Tell the AI who it is, where it works, and who it is talking to. This is the foundation everything else builds on.

Example: "You are a Senior Project Coordinator at a boutique interior design firm specializing in residential luxury. Your audience is high-end clients seeking clear communication."

2. Core Objective - The "What" Give the agent one primary job. A focused agent performs. An agent with ten jobs gets confused.

Example: "Your core objective is to create client update communications in our company voice, tone, and format."

3. Guidelines and Constraints - The "How" and "What Not to Do" Constraints are often more powerful than instructions. Telling AI what to avoid prevents the exact hallucinations and off-brand responses that eroded your trust in the first place.

Example: "Always separate emotional client feedback from physical design changes. Never invent timelines or budget figures. If key information is missing, ask before proceeding."

4. Tone and Voice - The "Vibe" If your whole team is using this agent to communicate with clients, it needs to sound like your firm, not write like a corporate lawyer. This is also where you can upload a reference document, like a tone and voice guide, so the agent absorbs your standards directly.

Example: "Warm, professional, and direct. Avoid corporate jargon and overly technical language. Write as if you are a trusted advisor, not a vendor."

5. Output Format - The "Look" Do not let the AI decide how to present the answer. Tell it exactly what the final product should look like so your team can copy, paste, and send.

Example: "Always deliver two sections: Section 1: a client-facing summary up to 3 sentences. Section 2: next steps listed as bullet points. Section 3: next meeting date and agenda"

6. Interaction Strategy - The "Loop" This is the move that turns your agent from a text generator into a thinking partner. Tell it to ask questions before it produces anything.

Example: "Before generating a response, review what I have shared. If the project name, deadline, or any context is missing, ask me clarifying questions before proceeding."

7. Examples - The Secret Sauce The fastest way to get AI to do exactly what you want is to show it what "great" looks like. In prompt engineering, this is called few-shot prompting. One strong example is worth a dozen extra instructions.

Example: Attach a past client update you are proud of and tell the agent: "Match this quality, structure, and tone in every output."

🎯 What This Looks Like Inside Your Business

Here are three places in your business where you can hire an AI agent today to support your team and feel the impact almost immediately.

Client Communications Consistent, clear, and concise communication is worth gold. Most clients are not fluent in your industry language.

A well built AI agent ensures that no matter who on your team sends the update, it reads like it came from the same trusted voice.

Clear. Professional. Easy to understand. Every time. In seconds.

Project Storytelling People connect with stories. The ability to translate a project approach into something a client can feel before a single decision is made is a skill not everyone on your team naturally has.

A well built AI agent allows anyone on your team to feed in the project details and walk away with a short, compelling narrative that deepens the client relationship and elevates the work before it even begins.

Internal Knowledge and SOPs How many times a week does someone come to you with a question you have already answered before?

A custom AI agent loaded with your standard operating procedures, your preferences, and your formatting standards can become the first place your team goes for answers instead of coming to you.

That alone is worth the setup time.

📌 Your Competitors Are Already Using AI Agents

While your team is manually drafting the fourth client update email this week, another business just like yours is sending that same update in a fraction of the time. In their voice. In their format. Without a second thought.

They are not smarter. They are not better resourced. They simply recognized a practical use case for what today’s technology can do.

Small gaps in efficiency like this compound quietly.

Fast, simple, and effective communication creates better informed and more connected clients.

These are the kinds of advantages AI agents create inside a business.

A poorly configured AI agent is still a bad employee. Vague instructions produce vague results.

The setup matters. That is exactly why this article exists.

🔄 Even Your Best Hire Needs Course Correction

Here is the part most people skip, and it is the difference between an impressive AI agent and a mediocre one.

When the agent produces work that is not quite right, do not just fix the output and move on. Fix the system.

Adjust the original instructions. Improve the sample documents. Clarify the standards you expect. The goal is not just better work this time. The goal is “better work” every time after that.

You can even ask the agent directly.

“Here is the result you gave me. Here is what I was hoping for. What should I change in the instructions so you can produce that next time?”

The answers often point exactly to what needs to be improved in the prompt.

This is what makes AI agents different from people. Once you teach them something properly, they do not forget. They do not drift. They do not revert back to old habits.

They simply keep getting better.

The only way they stop improving is if you stop refining them.

And that part is entirely within your control.

Figuring out where to start is usually the hardest part. If you want help building your first custom AI agent or identifying which task in your studio is the right first hire, that is exactly the kind of thing we love helping creative teams think through. Let's talk.