On ChatGPT for Product Managers

ChatGPT recently passed the final exam of Wharton's MBA program.

Well, most universities are considering this as bad news and downright unethical. Case in point — ChatGPT is banned from being used in exams in universities ranging from Harvard to MIT.

But on the flip side, it could be considered a positive, unstoppable force. Ethan Mollick, Associate Professor at Wharton, who is teaching students about innovation and entrepreneurship, has a more convincing rationale for moving forward. Simply put, 'If you can't beat them, join them'.

It would just be a matter of time before ChatGPT graduates summa cum laude in one of the toughest exams of all kinds (China's Gaokao exam?).

"I fully embraced AI for my classes this semester, requiring students to use AI tools in a number of ways. This policy attracted a lot of interest, and I thought it worthwhile to reflect on how it is going so far. The short answer is: great! But I have learned some early lessons that I think are worth passing on." — Ethan Mollick, Yale Professor

What would happen if we fully embraced ChatGPT for our knowledge work?

Product Managers on a typical day are juggling between conducting market research, defining roadmaps, managing stakeholders and prioritizing product features, among other things.

How might we make our lives as product managers easier using ChatGPT?

As humans, we have generally been apprehensive anytime something new comes up. Even if we look at the history of typewriters more closely, there was a lot of resistance. Bruce Bliven writes that when these machines were first introduced, "one real difficulty...was the public's feeling that typewriting, for private correspondence, was insulting, confusing or both".

Our predictions go wrong.

Not just for typewriters but even when it came to road transport. By the end of the 19th century, horses were predicted to fill the streets with poop. Even the Times had predicted that 50 years later, every street in London would be buried under nine feet of manure.

The situation is absurdly similar to how ChatGPT is now being looked at as a 'frightening tool' that cheats exams, spams dating apps, fool job recruiters and even 'steal jobs away from writers'. Many underestimate how ChatGPT can be used as a great thinking, writing, and everything-in-between tool. It's a copilot you never knew you wanted.

For the past few weeks, I've been experimenting with Midjourney and ChatGPT as a tool for thinking, writing, structuring content, paraphrasing or brainstorming ideas.

Let me share a few ways I've used ChatGPT in various contexts. (Although I'm not too fond of listicles, I will make an exception this time)

I want to cover some of the ChatGPT use cases for Product Managers/Builders in depth. But before that, I will first cover some of the most essentials which any knowledge worker in this era has to do — Reading and Writing.

On Reading

Part of my process for reading and digesting internet information is by asking guiding questions. For long-form articles which are 10,000 words+ in length, I sometimes make ChatGPT my study guide to digest these topics. I ask it to learn from it and suggest guiding questions.

Questions act as signages in your knowledge map. The best way to learn something new (and fast) is by asking questions.

I could ask ChatGPT to learn something from an article online quickly.

And then, I prompt:

Ask me 10 questions about the article I provided

As we finetune the prompts, the questions get better over time. This works magically not just for online blogs but also for research papers, books, newspapers, and legal docs. Anything you would need.

Filechat.io, for example, lets you upload docs of all kinds and lets you ask questions through a personalised chatbot.

Filechat

Filechat is the perfect tool to explore documents using artificial intelligence. Simply upload your PDF and start asking questions to your personalized chatbot.

Through ChatGPT, anything could be converted into a chatbot.

On Writing

When I first started, I tried something very generic as follows —

This was giving very generic, textbookish material not worthy of usage anywhere. It was meh at best.

That's where prompts come in.

We could steer the ship in any direction.

For example,

Write an essay with the following points. Use a business tone. Use at least two clear examples. Make it concise. Write for a well-informed audience. use a style like the New Yorker. Make it at least 7 paragraphs. vary the language in each one. End with an positive note. - Small risks in healthcare can prove to be dangerous - Data in which AI is trained might not be of good quality -In complex systems, some errors are catastrophic - Future trends are promising

Another specific use case for product managers is fine-tuning the writing style in various personas.

You can ask ChatGPT to talk to you as a journalist. A travel guide. A journalist. Or even as a character from a movie or a TV Show.

Storyteller. Advertiser. Relationship Coach. Motivational Speaker. And various other personas. You can find the entire list here.

And finally, even as a Product Manager.

Now, coming to the more product stuff.

Drafting Product Specs

I recently tried prompting it to generate product specs for a Product I was working on.

Write product specs for an Whatsapp-based healthcare chatbot service for a healthcare NGO based in Bangladesh that allows mothers to receive information on healthy behaviors through personalised support and in-time reminders. This format should be followed

Summary

• User Need/Problem
• Business Metrics
• Primary Research (User Interviews, User Testing)
• Secondary Research (Competitor Analysis, Benchmarks, Surveys)
• Features in the form of Jobs-to-be-done
• Post Mortem

Not that it was highly accurate. I had not described the detailed product information yet. But it was able to pick up enough juice to work with.

ChatGPT was proving to be a good first-drafter. No more staring at blank spaces!

Designing Interview Scripts

Coming to Interview Scripts.

During the early phase of defining our product, we spoke to various caregivers from different divisions within Bangladesh to understand their healthcare journeys. Do they have trust in the public health system? Where is their first touchpoint to access health information? We had so many questions.

We had done this all manually. And ChatGPT didn't exist back then.

If I were to do it now, I would use this prompt to generate an interview script.

Create non-leading user interview questions for <product/feature description> that passes the Mom Test (book by Rob Fitzpatrick)

Defining Metrics

Next, I wanted to explore another crucial part of product management — Defining metrics.

In my case, I was looking to define metrics for our healthcare chatbot platform. Let's see what ChatGPT thinks.

And If I had to boost the engagement rate specifically and was looking for possible design methods I could use to fit this application, I tried this prompt —

What are the various design methods I could use to test the engagement of the overall service? You’ll only reply me the text-based 10 rows excel sheet with row numbers and cell letters as columns (A to B). Column A should indicate 'Design method', and Column 'B' should indicate how I could use that design method for my application. Any methods by IDEO or Nielsen Norman Group is preferred.

Not bad. Cohort analysis to track the engagement metric is not a bad suggestion. I could go into any of these rabbit holes and ask ChatGPT to create a full-fledged experiment on Cohort Analysis.

Design an experiment to track engagement metrics over time for specific groups of mothers to identify trends and patterns using Cohort Analysis. Give the answer in the form of an excel template.

This was surprising. Not only did it give me an excel template, but it also gave me step-by-step instructions on how to conduct the cohort analysis and how it could be measured.

Business Viability

Various prompts could help you understand the viability of your offering.

For example, here, I use the Business Model Canvas through this prompt —

Please create a simple business model canvas for the healthcare chatbot service in a table format

Or figuring out a go-to-market strategy. Here I wanted to understand how this healthcare chat service could reach up to 1M users in 2023 for a different geography — Indonesia.

As shown earlier, we can induce various personas to provide answers. I invoked the persona of a CEO who is looking at global expansion. ChatGPT remembers my previous queries, so I wanted to ask how the non-profit could expand from Bangladesh to Indonesia.

The replies were a bit generic, and I was not so satisfied.

I fine-tuned the prompts a bit more by using these addendums —

Not bad. It was now giving me concrete names. Instead of just saying reach out to any NGO, it told me to contact Yayasan Kusuma Buana Foundation.

Instead of telling me to form partnerships. It told me to connect with the Ministry of Health, PKBI, and Cipto Mangunkusumo Hospital.

Now, that was something.

Wow.

On Branding

While road mapping the target goals for this year, we divided our larger product/service into smaller chunks (projects) for easier operations.

I was stuck a bit with finding appropriate names for the projects. It had to convey the essence and had to be short and sweet. I turned to ChatGPT to find some brandable names.

Again, the first list of names was pretty generic.

So I gave some reinforcing prompts to steer the responses. I injected some more requirements, which made the responses even better.

This surprisingly gave us good results. We ended up naming our telesupport platform 'Speedy100'.

Drafting customer journeys

Next up was customer journeys.

While describing the overall customer journey of the mother accessing the healthcare chatbot service, I wanted to bring in much more structure. I wanted to segregate the key phases into touchpoints, jobs to be done, potential pains, potential gains etc.

I described the customer journey and asked ChatGPT to outline it how I wanted.

Here's what I got.

After trying out all these use cases, I realised this is just the tip of the iceberg.

There are so many more ways in which ChatGPT could empower us.

But I will pause right now to have some brevity.

The next three editions will cover most of the ChatGPT applications and various examples of how I use them in my day-to-day work.

Subscribe for more updates :)