Week 3 - Friday 9th May.
Post integrating Google CSE (to enhance news coverage & quality), I focused on ensuring the Linkedin Drafts are as close to a user's own style as possible.
Achieved this by tapping into the System-Prompt of genAI. This is where we define things like
"Write with conviction and strategic clarity. Start with a bold claim or insight. Keep the tone confident, concise, and jargon-smart. Use short sentences. Prefer decisive language ("we must", "it's obvious", "the data says"). Focus on ideas that drive action or shift mindsets. Avoid hedging or qualifiers. Avoid emojis and small talk."
Considered two possible approaches
- Upload a few of your linkedin posts and the AI decodes your writing style
- Choose from one of the pre-defined voice formats and edit it to suit your style.
Chose to go ahead with the latter.
Why? It seemed like a lower friction option (pulling old posts and uploading on platform is some effort) for users and was also an easier development for me.
So what you will be able to do now - is choose a voice - predefined or customized (upto 3 ) of your own to craft the final Linkedin Post.
I have been testing it and quite amazed by the variety of final Linkedin Posts it generates by just toggling across the voices.
Give it a shot and as always - I am keen to hear your feedback and inputs on how to improve it further.