So after all this time, you have given in. You have registered for ChatGPT, perhaps purchased a Jasper subscription,n or started using Claude. Honestly? That was a little magical at first. You asked it to create an email, and suddenly, there it was.
However, something changed.
The emails began to sound unnatural. A little too robotic. A little too generic.
What is not told about artificial intelligence tools designed for business they are not psychic. They are more like new college graduates who want to make a good impression. They have tremendous potential. However, they require training.
Let’s get started today.
Why Your AI Sounds Like a Stranger
You have hired a new employee and handed her a laptop and pointed to a desk, saying, “OK, you may now send e-mail messages to your customers.” No additional information is provided about the company or the tone of the e-mails she will write. What happens?
She sends e-mail messages that are completely robotic-sounding because that is exactly what she was instructed to do. That is essentially what we all do when we use AI. We create a chat window and magically expect results. And instead, we receive generic responses.
All business-related AI software has a default generic mode. This is true because the training data used by these programs includes the entire Internet. Therefore, it has knowledge of a great deal about virtually every subject, but also knows very little about anything specifically related to that subject. Generic input will always produce generic output.
The good news is that you can easily correct this issue in about one-half day.
Step 1: Feed It Your Greatest Hits
There is an odd fact about training A.I.: It uses the same method as humans in learning by example. Show it ten of the best e-mails you have ever created, and it will begin to know your style.
Begin with a brain dump document.
Open a newGoogle Docc or Notion page. Name this document, “A.I. Training Document,” or whatever name sounds most formal. Begin writing.
The document that you create here is the base of every single thing your A.I. generates.
| What to Include |
|
|
| Brand voice | Friendly but professional. We use contractions. We keep it real. | |
| Best emails | 5 emails that got replies or sales | |
| Top content | 3 product descriptions that convert | |
| Customer FAQs | 5 most common questions | |
| Banned words | leverage, synergy, utilize, circle back |
Then do something simple. Upload that document to ChatGPT, Claude, or whatever tool you’re using. Say: “This is my brand voice. Read this before you write anything for me.”
Suddenly, your AI knows you. It’s not guessing anymore.
Step 2: Create Prompt Templates (Your Secret Weapon)
START_TEXT
Let me take a guess. Every time you start using ChatGPT, you spend thirty seconds staring at the blank box, trying to remember what you asked it last time.
Most people give up right there. They get mad at the AI tool. The issue isn’t the tool; it’s the lack of a system.
Save your prompts as templates. Use them like recipes.
Here is a complete template you can copy and paste today:
“Write for me as my in-house content writer. Please write in a conversationally written (founder-to-founder) tone. I want you to avoid buzzwords completely. Write no longer than two sentences per paragraph. Begin by stating a real business pain point that your reader will be able to relate to. State a single, actionable takeaway that they can apply to their business today. Keep all responses under 300 words unless I ask you to do otherwise.”
Use this template for blog intro writing, drafting an email, creating a LinkedIn post, etc.
Save this. Use this. Save this again. Use this again.
Step 3: The Art of “This Is Almost Right”
The first time someone completes something well is never the first attempt at completing something well. Humans and AI both know this. The magic in working with an AI occurs in the give and take of the back-and-forth process.
One way to get into a better mental space when using an AI is by thinking about the AI as your drafting partner rather than as the finisher of a piece of work. Your role is to provide feedback; the role of the AI is to take another shot at the work.
A sample of how the feedback loop would flow is:
You ask for something the AI gives you something.
You respond with: “That is a good start; however, I would like to see it shortened. Remove the fluff. Start with the punch line.” The AI attempts again.
You respond with: “better; now include a specific example”.The AI provides the next iteration. Three iterations of the same process can produce something useful. maybe even something great.
Pro-tip: When the AI produces something that is successful, be sure to let the AI know why it succeeded. Tell the AI something such as: “I liked the last version so much because you were able to keep it under 100 words and write in a conversational style”. Praise helps an AI learn, just like it does for people.
Step 4: Build Your Custom GPT (If You’re Serious)
Once you’ve established a good workflow and are comfortable working with ChatGPT, you can take the process to the next level by setting up Custom GPTs (if you’ve purchased ChatGPT Plus). A Custom GPT is essentially your own personal version of ChatGPT that has been trained on all the information you provide it, including what you tell it during the training phase. This means it will remember your style, voice, preferences, and even your writing speed.
Custom GPTs allow you to train a version of ChatGPT to be much faster at drafting content and require fewer revisions than you would otherwise experience.
Creating a Custom GPT takes approximately 10 minutes.
Here’s how to do it:
- Open ChatGPT.
2. Click “Explore” or “My GPTs” in the left-hand sidebar.
3. Click “Create a GPT.”
4. Give it a name: “Growrai Writer”, “Marketing Assistant”, etc.
5. In the instructions section of the dialog box, insert your Brain Dump Document (the document containing your thoughts, opinions, experiences, values, strengths, weaknesses, likes, and dislikes). In addition to telling the GPT who you are and how you write, include what you dislike most about how others communicate (jargon, verbosity, etc.) and what you enjoy most about communication (succinctness, clarity, etc.).
6. Upload three to five example documents (email, blog post, social media post) to help the GPT learn your tone and style.
7. Provide the GPT with instructions on how to conduct itself when responding to questions. For example: “If ever unsure of anything, always ask for additional details.” “Never speak in corporate jargon.” “All answers should begin with a concise summary.”
When you have completed these steps, you will now have a customized version of ChatGPT that is specifically designed to know and understand you. Every time you open the Custom GPT, it will retain the knowledge it gained from previous conversations, so you don’t have to repeat yourself.
Important Safety Note: Do NOT upload confidential documents such as contracts, client lists, or sensitive financial data. When possible, upload anonymous examples instead. Remember, your data belongs to you alone!
The creation of a Custom GPT is the key differentiator that separates those individuals who merely dabble in AI technology from those who successfully apply AI technologies for business purposes. The cost of creating a Custom GPT is approximately $20 per month, and once created, it will save you literally hundreds of hours of work in the future.
Step 5: Train It on Your Customers, Not Just Your Brand
Most of the writing training happens here, and that’s where it should stop.
Your brand voice is important; however, your customers’ voices are pure gold.
If you want to create an AI that will write content people will actually respond to, train it with your target audience. Train it using a combination of the following:
- Anonymized Customer Support Emails.
- Reviews written by your customers.
- Comments made in response to your social media postings.
- FAQ responses that actually answered the customer’s question.
Do not feed it raw data; curation is better than the amount of data. Every time, quality examples beat quantity of examples.
Why? Because at this point, the AI not only knows how you speak, but also knows how your customers speak. The AI understands their concerns, their frustration, and their exact wording.
Write a prompt such as: “Read these 10 customer reviews for my business and describe to me the patterns you see, what they like, what they find confusing, what words do they use?”
Use those insights to help develop everything else.
Step 6: The Maintenance Schedule (Yes, You Need One)
The fact about training most people are uncomfortable with is that it does not occur just once, but as an ongoing relationship. It requires attention.
Your business is going to evolve; your products are going to evolve; your target audience is going to evolve; and therefore, so will your artificial intelligence (AI).
Schedule yourself quarterly to do the following:
- Update your training documentation with new examples
- Remove outdated information that no longer applies
- Add new voice guidelines if you have shifted your focus. Test the prompts you have saved in your database — do they still function?
- Spending 15 minutes each quarter on this will keep your AI current.
Real Examples: What Good Training Looks Like
To demonstrate the difference that training can make on an AI’s generated message;
Training before sending the message:
“Dear Customer, we hope this email finds you well. We wanted to contact you to let you know about our new product line, which will have innovative ways to meet the requirements of your business.”
After Training (with actual emails sent by a coffee roaster):
“Hi [Name], we’ve recently roasted something we think you’ll really enjoy, a small-batch, limited quantity coffee from a small farm in Colombia. It is bright, it is fruit-flavored, and there are only twenty available to purchase. As always, first choice goes to you.”
The difference, as stated above, is that one is corporate-sounding, while the other is friendly-sounding.
This is what training does: it doesn’t make an AI smarter, but rather makes it seem like you.
The Tools That Make Training Easier
“You do not need to have all of those. The best tool for the job is one that has been properly trained; five poorly used tools will never be as effective as one that has been properly trained.”
When discussing AI tools in a business setting, this is where the most useful training can occur:
| Tool | Best For | Training Feature |
| ChatGPT Plus (Recommended) | General writing, brainstorming | Custom GPTs, memory, file uploads |
| Claude (Recommended) | Long documents, analysis | Project knowledge bases, huge context windows |
| Notion AI (Recommended) | Internal docs, wikis | Learns from your workspace content |
| Jasper (Optional) | Marketing copy | Brand voice templates, style guides |
| Copy.ai (Optional) | Social media, emails | Infobase (saved knowledge about your brand) |
| Perplexity (Optional) | Research | Remembers your past queries and preferences |
Choose one primary tool and use it consistently. Continuously jumping from tool to tool will confuse you more than provide value.
What Training Won’t Fix (Be Realistic)
We are being truthful for one minute now. Training solves a lot of things; however, not all of them.
Training will not solve:
- Poor strategy (no AI can help with a poor concept)
- No edits (the final product requires human eyes to view it)
- Data privacy problems (be cautious about what you are uploading)
- Overly high expectations (it is an aid, not a magic wand)
- If there is something wrong with your business model or offering, training cannot repair it. Training makes something good even better. It does not transform a poor offering into a good one.Do not have overly large expectations of technology, as you will likely become unfulfilled with the results.
Your Seven-Day Training Plan
START_TEXT
Want a quick way to begin? This is your week:
Day 1: Begin your Brain Dump Document. Collect 5-10 examples of what you consider “good” writing.
Day 2: Design your first Prompt Template. Choose something you perform weekly as an example.
Day 3: Test it. Run the prompt, get output, provide feedback, and make changes.
Day 4: Create a Custom GPT (if you are on Plus), or create a Saved Prompt System.
Day 5: Use Customer Data reviews, email, FAQ’s to feed it.
Day 6: Produce three things using your Trained AI. Compare to the previous output.
Day 7: Remember to check in on your Quarter. That will mark completion.
You do not require perfection. You require momentum.
The Bottom Line
AI has become ubiquitous today. If your competitor is utilizing AI, you have the option to do the same; however, you will be untrained, while they will be trained.
- Trained AI allows your competitors to compete as themselves, while untrained AI makes them all sound alike.
- In a world of sameness, being unique is the greatest competitive advantage.
- Open a new document right now and paste an actual email into it. That is sufficient to start.
- The very best business AI tools are not those with the most features; they are those that know you better.