This week's tip: Reduce complexity when drafting a new workflow for AI.
When you have an idea for a new AI workflow, the best way to go test it is with an AI MVPT (minimum viable prompt).
While the MVP (minimum viable product) is common at most companies, you should follow this process as an AI adopter as well for testing new AI workflows.
// And if this MVPT thing sounds too fancy for you, just think of it as a draft.
Here are the things that matter when launching your AI MVPT:
1. Clear objective: What task or problem are you looking to solve with AI?
2. Simple prompt: A straightforward prompt that clearly communicates the goal.
3. Evaluation criteria: Decide on how you will judge the success of the output.
That's it.
To see an example of how I develop MVPTs, let's look at an idea for a project management workflow I had:
I wanted to automate the process of generating status reports for ongoing projects using ChatGPT. The goal was to save time and ensure consistency in reporting.
With this prompt, I wanted to prove or disprove a hypothesis that ChatGPT could accurately summarize project updates and tasks from a list of bullet points.
So I crafted a prompt that included a list of project updates and asked ChatGPT to generate a concise status report. I provided examples of previous reports for context and specified the tone and format I was looking for.
In a short time, using my MVPT Method, I proved that ChatGPT can handle the workflow and moved closer to completely delegating it.
So, skip the time-consuming and fancy prompt engineering techniques.
Start by proving that ChatGPT will actually handle your workflow.
Refine it from there. Test, iterate, test, and improve as you go.
Get results. You can always complicate things later.
Here's how you to structure the actual ChatGPT prompt to test the workflow:
To stay informed about AI for Business:
Follow & Subscribe: LinkedIn & Newsletter
For feedback, questions, ideas Contact Me
Comentarios