Understanding Lightweight AI Assistants
Before customising, it's important to know what a 'lightweight' AI assistant is. Unlike heavy-duty enterprise solutions, these are often extensions, plugins, or integrated features within the tools you already use, like Gmail or Outlook. Think of assistants
like Gemini in Gmail, Microsoft Copilot, or other third-party tools that focus specifically on email tasks. Their goal isn't to replace your thinking, but to handle repetitive work like drafting replies, summarising long threads, and filtering your inbox. The key advantage is that they layer onto your existing workflow rather than forcing you to adopt a whole new platform.
Define Your Authentic Writing Voice
You can't teach an AI your style if you haven't defined it yourself. Before touching any settings, analyse your own writing. Are you formal or conversational? Do you use humour? Are your emails concise and direct, or more detailed and explanatory? Look through your sent folder for 5-10 emails you feel perfectly represent your professional persona. Note your typical greetings, sign-offs, sentence structure, and even your preferred emojis. This 'style blueprint' becomes the foundation for all your customisation efforts. Without it, the AI will default to a generic, robotic tone that feels inauthentic.
Mastering Custom Instructions
Most modern AI tools, including ChatGPT and its email-focused cousins, feature a 'custom instructions' or 'profile' section. This is the most powerful tool for personalisation. Here, you provide permanent guidelines that the AI will remember for all future tasks. Instead of repeating yourself with every prompt, you set the rules once. Use this section to tell the AI your role, your company, your target audience, and most importantly, the writing style you defined earlier. For example, you can instruct it: "Act as a marketing manager. Write in a friendly but concise tone. Keep emails under 150 words and always end with a clear call-to-action."
Provide High-Quality Examples
AI models learn from examples. The fastest way to train your assistant is to show it what you want. Many tools now allow you to upload documents or paste text to help them learn your voice. Paste in those ideal emails you collected earlier and use a prompt like: "Analyse the following emails to understand my writing style. Summarise my tone, structure, and common phrasing." The AI can then create its own style guide based on your actual writing, which is often more effective than just describing it. The more high-quality, representative samples you provide, the more accurately the AI can mimic you.
Develop Reusable Prompts and Templates
Beyond general instructions, effective customisation involves creating a library of specific prompts for recurring tasks. Prompt engineering is the practice of designing inputs that get the best possible output from an AI. For instance, instead of just asking the AI to "write a follow-up email," create a detailed prompt: "Draft a polite two-paragraph follow-up to a client who hasn't responded to a proposal sent three days ago. Reference the proposal's main benefit and suggest a brief call next week." You can save these detailed prompts as templates, ensuring consistency and quality every time.
Refine Through Iteration and Feedback
An AI assistant is like a new intern; it needs training and feedback to improve. The initial drafts it produces won't be perfect. Your job is to treat them as first drafts, not final products. When an AI-generated email misses the mark, don't just delete it. Edit the text to match your voice, and then feed the corrected version back to the AI with the instruction, "Use this edited version as a better example of my tone going forward." This iterative process of generating, editing, and reinforcing creates a powerful feedback loop, making the assistant smarter and more aligned with your style over time.
