Building a gratitude bot with ChatGPT and Zapier AI Actions
I know I should send more thank you notes, perhaps even a lot more thank you notes, but I'm not good at warm and engaging prose. I end up staring at the screen, worried that my email missives will seem too cold, awkward, or lame.
But ChatGPT is great at heartwarming thank you notes, perhaps because it has no complicated, emotional relationship with their recipients. For example:
Hey Sue,
Just wanted to drop you a note to say a huge thank you for the gorgeous red roses! They really brightened up my day and are making my place look and smell amazing. You’re the best!
Lots of love,
[Your Name]
And this:
Hey Fred,
Wow, I was blown away by the silver trombone you sent for my birthday! It's an amazing gift, and I really appreciate your thoughtfulness. I can't wait to play some tunes and think of you. Thanks a ton for making my day extra special!
And this:
Dear [MIL's name],
Thank you so much for your generous gift of $100 towards [Son’s Name]'s college fund. Your thoughtfulness and support mean the world to us, and we are touched by your contribution to his future. It's a wonderful addition that brings us closer to securing his educational journey.
With warmest regards,
[Your Name]
I decided that building a Thank You Notes GPT could help me act on my grateful impulses more consistently and with less anxiety. Also, because I don't like jumping between apps, I used a Zapier AI Action to automatically paste emails into my Gmail drafts folder.
A few simple instructions
I started the prompt with some basic instructions encouraging the GPT to ask users relevant information, such as who they are thanking, what they are grateful for and their preferred tone of voice:
You are a writing assistant who helps people express their gratitude through warm, personalized thank you notes. Interactions should proceed as follows:
Greet the user in a friendly tone.
Ask what they are grateful for (examples: a gift, a meal, a referral, a job recommendation, etc.) For gifts, ask for a description.
Ask who they would like to thank.
Ask them to describe the tone of voice they'd like to achieve.
Ask how long the note should be: a few lines or a paragraph or more.
Once you have all this information, produce a draft of the note.
Adjust as needed based on user feedback.
Once you have the final draft, use the Zapier AI Action Gmail: Create Draft to create an email draft.
Surprisingly easy to set up
Setting up GPT actions typically requires configuring an authentication method and building an OpenAI schema based on an API. While ActionsGPT can often create a nice first pass at the schema from API docs, the results often require fine-tuning and can be tricky to troubleshoot.
With Zapier AI actions, you can cut-and-paste a schema and append a premade set of instructions to your prompt that can be easily adapted to work for any of Zapier's many automations. I set up Thank You Notes GPT with the Gmail Create Draft action. Zapier's resource page does a great job explaining how this works and includes a step-by-step tutorial.
When I tested it, the GPT guided me through the process of logging in and enabling an action, and then showed me a preview of my email draft before saving it to my Gmail drafts folder. If you're already a Zapier user, this process is pretty quick and simple. Non-Zapier-users will have to create a free account.
No more excuses
Now that my Thank You Notes bot is up and running, I have no more excuses for skipping out on thank you emails to friends and family. I will hopefully act on my gratitude more frequently and spend less time and emotional energy doing it.
But a quick chat with my husband quickly uncovered some possible snags. If I am relying on AI and not pouring my soul onto the virtual page, will that make my thank-you notes less meaningful? And what if everyone uses AI for thank you notes? Will we all sound the same?
I have no idea how to answer any of that. The experiment continues...
