Why Default AI Responses Fail
Out of the box, large language models (LLMs) like ChatGPT default to a neutral, risk-free voice designed to be broadly applicable. This results in emails that can feel flat, overly formal, and disconnected from your personal brand. When every response
sounds the same, you lose the unique voice that builds rapport with colleagues and clients. The goal is to move the AI from a generic intern to a trained assistant that understands nuance, saving you time on tedious edits while preserving your professional identity.
Step 1: Gather Your Writing Samples
The first step is to provide the AI with high-quality examples of your own writing. You don't need a massive archive; 5 to 10 representative emails can be enough for the AI to detect consistent patterns. Export sent emails or gather documents, blog posts, or even casual messages that reflect your authentic tone. The goal is to give the model concrete examples of your typical sentence structure, vocabulary, and level of formality. Some tools can even analyze your email history automatically to create a personalized tone model.
Step 2: Create a Personal Style Guide
Once you have your samples, the next step is to create a simple style guide. This involves translating your writing habits into a short list of clear rules for the AI. You can even ask an AI tool to analyze your samples and identify your key patterns for you. Your guide should define your foundational tone—are you warm and conversational, or formal and direct? It should also include specific language patterns, such as preferred greetings ("Hey team" vs. "Dear all"), sign-offs, and common phrases you use. Finally, specify structural preferences, like your use of bullet points, bolding, and typical paragraph length. Explicitly stating what to avoid, such as corporate jargon or emojis, is just as important.
Step 3: Master the Art of the Prompt
Effective prompting is crucial for getting the desired output. Instead of a vague request, structure your prompt with clear instructions separated by delimiters like '###'. Start by assigning the AI a role, such as, "You are my professional email assistant." Then, provide your style guide or key rules from it. Finally, state the specific task. For example: "Using my style guide, draft a polite follow-up email to a client about the pending invoice, keeping it under 100 words." Providing a few examples of the desired output within the prompt can further refine the results.
Step 4: Use Persistent Custom Instructions
To avoid pasting your style guide into every prompt, use features like 'Custom Instructions' in tools such as ChatGPT. This feature allows you to save information about yourself and your preferred response format, which the AI will then consider in all future conversations. In the first field, you can describe your role and company context. In the second, you can outline your writing style rules, such as tone, length, and formatting requirements. This creates a persistent personality for the AI, making it a more efficient and personalized assistant over time.
Step 5: Iterate and Refine with Feedback
Training an AI is an ongoing process, not a one-time setup. The initial drafts it produces are starting points, not final products. The real learning happens when you provide corrective feedback. When an output misses the mark, don't just edit it; tell the AI what was wrong and why. For instance, you might respond, "This is too formal. Rewrite it to be more conversational and use shorter sentences, like in the examples I provided." This iterative loop of generating, editing, and providing feedback is what truly sharpens the AI's ability to mimic your voice accurately.
