How to Turn ChatGPT into Your Personal Consultant: A 5-Step Approach

Article written ~98% by me, 2% ChatGPT.?

If you're like me, you've read a lot about best practices for prompt engineering. You've likely found articles that share various acronyms and quirky tips that tell you how to engage with ChatGPT, focused especially on that first prompt - how to lay out your request in one perfect message that gives you the response you need. Then you open up a new chat and you stare at the blank screen, intimidated and uncertain. Or worse, you spend far too long wording and rewording your initial prompt, only to be repeatedly disappointed by the results.

Rather than sharing yet another prompt engineering guide, I want to provide something different - a framework for how to use ChatGPT as your personal consultant and coach on any topic, regardless of your expertise in it.

The Five Steps

1. Set the Stage?(optional but highly recommended): This can mean inputting a document as a knowledge base or prompting for a summary of principles or best practices in your domain of interest.
2. Script: In this step you will prompt for a set of open-ended questions, or in other words an interview script.
3. Share: Ask ChatGPT to serve as an expert in your domain and use the script to interview you. Instruct it to follow up on questions as needed. Reflect and share your responses as you go through the script.
4. Summarize: After reaching the end of the interview question set, ask ChatGPT to summarize the key insights and takeaways. This step not only helps you review and reflect on the conversation, it also serves to reset the most recent context with the key points from your conversation.
5. Strategize: Finally, request ChatGPT to propose a plan of action or suggestions, and adjust it to your liking.

Goal-Setting Use Case

Here's an example from my work life. I've never liked goal-setting, so when my team was tasked with creating SMART goals for ourselves, I saw it as the perfect opportunity to enlist ChatGPT. Here's how I applied the approach. Each step begins with the prompt I used:

1. Set the Stage:

Review this tracker, which contains instructions for developing SMART goals as well as example goals from 2 of my teammates. Provide a summary.        

I uploaded the template my teammate created, which conveniently already had 2 team members' goals filled in. This was a great example of few-shot prompting that required little effort on my part. I asked ChatGPT to summarize the document.

This step helps ChatGPT “get into the mindset” of an expert in the particular domain where you need help. It takes advantage of a characteristic of LLMs called the context window, which is the text its using to predict its next response (since AI is like advanced autocomplete, after all). By setting the stage with context, I’m giving myself some control over the AI's training data, in a sense. As an alternative to this, if you don’t have any documentation to upload, you can prompt for a list of guiding principles or best practices in the domain you need help in.

2. Script:

Acting as a professional coach, create a list of interview questions that can be used to help someone reflect on their role and ambitions and generate ideas for goals.        

Staring at a blank chat is intimidating, and prompting something like "help me create SMART goals" doesn't provide enough context specific to you for useful responses. You need scaffolding: a structure to draw out the context that ChatGPT needs from you to give you high-value responses. A script provides this for you. If you're familiar with the expression "help me help you" - this is like helping ChatGPT help you…to help you.

So I asked for interview questions, and it created a list of ~25 questions - too overwhelming. I didn't bother reviewing the questions, I just asked for the 5 most important. I could’ve asked for 5 questions from the beginning but having that long list as additional context setting actually improved my later responses.

3. Share:

Perfect. Now please guide me through a conversation using the question list we just came up with. Ask one question at a time. After each question, return to the question list and consider whether we should proceed with the next question or insert a follow up question to dive deeper into the topic that has just been discussed.        

Next we started the interview. I provided my responses and was quite surprised at how well my "interviewer" balanced the task of working through our question list and following up on my responses to draw out more detail. Here's an example

This step is your opportunity to reflect and further tune that context window that you set, giving your consultant information to make suggestions tailored to you. The key here is to abide by "garbage in, garbage out" - do your best to answer questions and follow ups thoughtfully, because the effort you put in will pay off in a better final strategy.

4. ?Summarize:

Let's move on from the questions and transition into generating goals that I can input into the tracker. First please summarize our conversation so far.        

After the final question in our list ChatGPT tried to ask a follow-up again, but I felt I was done reflecting so I prompted for a summary and reviewed our conversation.

5. Strategize:

Please suggest 5 SMART goals for me in light of our conversation.        

Finally I asked it to suggest 5 goals for me based on our conversation. I really only wanted 2, but I knew I wouldn't want to use all of its suggestions, and ChatGPT excels at idea generation. Five felt like the right mix of not giving myself too much to read through and evaluate (OK, I started to get lazy after all that reflection), but enough to have some good options. I ended up using one suggestion almost verbatim and taking the general idea of another but adjusting it to a different context.

So this step still takes some work from you, but it’s because strategizing is a skill where humans still have a decent leg up on AI. The huge benefit of this approach is it helps you reflect and synthesize the information you need to do what's best for you.?

As you can see from the example, this framework leverages the strengths of ChatGPT to elevate your decision-making. It's an opportunity to move past the paralysis of staring at a blank chat window or the futility of iterating on the perfect prompt. Using this approach, I was able to reflect and strategize more efficiently and effectively, not only making a dreaded task less painful but also ultimately arriving at a better outcome.

?

I want to extend a huge thanks to Jared Valdron for encouraging me to write this article and providing great feedback to make it better!