AIforWork.co Makes Complex Prompt Engineering Easy
Saren Sakurai
15+ years driving growth through marketing innovation. Expertise in engagement, persuasion and scaling business performance. Proven success orchestrating marketing strategies and turning insights into measurable results.
Prompt Engineering is a fast-rising, in-demand skill across every department and discipline. Whatever you do, the better you get at writing prompts, the better results you'll get back, the better you'll be able to integrate AI into your day job. Setting the right context is important to center the conversation on a specific topic so that the agent can hone its answer to be just what you need.
AIforWork.co is a web app with a seemingly simple premise, to provide prompts in the context of your role at work. Choose a department, then select a job title, and you'll get a list of options like "Create An Email Content Calendar" or "Create a GDPR Compliance Checklist." All good, makes sense. Plenty of sites provide something like this. What I think sets AI for Work apart is the detailed engineering within the prompts you copy and paste into ChatGPT-4.
Upon entering one of the provided prompts, ChatGPT-4 responds automatically with five key questions to refine its forthcoming answer. When you answer the questions, you'll see its first draft, but you'll also receive a ChatGPT-4 self-evaluation made on 6 dimensions: Accuracy, Completeness, Clarity, Use of Reference Material, POV from Industry Expert, and an Overall Rating. Each will receive a score from 1 to 10.
You can accept the first draft as good enough, but ideally you'd respect the loop and start to iterate with ChatGPT-4. The loop prompts you to revise the draft, each draft is annotated with reasoning behind decisions, and each round is re-scored. You can repeat this loop until the scores approach 10. Every prompt is evaluated against published books on the subject at hand. For example, most of the copywriting tasks are scored based on Strunk & White's Elements of Style. Each of the 2,000 prompts is checked against 3 specific books from hthat specific field. All of this detail is programmed into the original prompts themselves. It's really good.
Here is a video explaining the AIforWork.co "Task + Evaluation" framework:
领英推荐
Here's where I think you'll find the secret sauce when actually using these prompts at work: as you loop through the prompt chain, ChatGPT-4 will annotate the prompts with its reasoning behind each of its responses, and it will cite the text from which the best practice was drawn. This reveals exactly how it arrived at the suggested result AND established the Why it chose to respond the way it did. The brilliance, I believe, is that as an employee utilizing the results of this loop in your real-world job, you're provided with a footnote that you can conveniently cite at the bottom of your presentation, from a established author or text, known as an SME. It's quite a cheat code for producing higher-level work, without actually doing the homework.
Give AIforWork.co a try and let me know how the prompts work for you. Don't settle for the first response, loop it a couple times before releasing to the wild. I'd love to hear how it works for you in the real world. Comment below.
Building AI Agents ??
1 年Hey Saren Sakurai thank you for sharing about AI for Work! Great post! ??