ChatGPT is a language model developed by OpenAI designed to understand and generate text fluidly and consistently. For a Product Manager, ChatGPT can simplify a wide variety of tasks, from the most creative to the most repetitive, and even the most daunting (summarizing user interviews)! In this article, our Head of Product Data & AI Sébastien Altounian gives all the keys to using ChatGPT in his daily life as a Product Manager.
I- When to use ChatGPT as a Product Manager?
- Writing Users Stories
- Analyzing and synthesizing data
- Drafting market research
- Preparing user interviews
- Transcribing user interviews
- Formulating the problem statement
- Finding new ideas
- Writing content
- Mapping your product's UX
- Making an impact study
- Identifying relevant KPIs
II- How to create a good ChatGPT prompt for Product Manager?
III- The best ChatGPT prompts for Product Manager
- Writing Users Stories
- Analyzing and synthesizing data
- Drafting market research
- Preparing user interviews
- Transcribing user interviews
- Formulating the problem statement
- Finding new ideas
- Writing content
- Mapping your product's UX
- Making an impact study
- Identifying relevant KPIs
IV- Use of generative AI by Product Managers: what about ethics ?
When to use ChatGPT as a Product Manager?
Here's a non-exhaustive overview of the main use cases of a generative AI like ChatGPT for a Product Manager. Some of these use cases are largely automatable, leaving the Product Manager with only residual tasks of proofreading, refining or interpretation. In other cases, a ChatGPT-type AI serves more to initiate or structure thinking: it's up to the PM to enrich this base at a later stage.
We'll assign each of these use cases an automation score from 0 to 3 below, corresponding to the following scale. Please note that even minimal human validation is always required!
This automation scale allows you to rank your features and tasks according to the degree of human intervention maintained or removed by the addition of artificial intelligence.
Writing User Stories
Tired of churning out User Stories (US)? Once ChatGPT has been introduced to the best practices in this field, it can save you precious time by writing precise and complete user stories for you.
Automation level: 2/3
Type of requirement: delivery and copywriting.
Kevin Si, Product Manager at Thiga, uses ChatGPT to help him write US. "Sometimes there are a lot of elements to add to the user story, especially if you have a lot of business rules specific to your company. In other cases, the result is almost perfect, as when you're doing a minor PHP version upgrade. In any case, it's a great help, because it means I don't have to start from a blank sheet of paper, I just need to tweak just a few pieces of information. However, be sure to obtain your employer's permission before using ChatGPT!"
Analyzing and synthesizing data
You've collected a large amount of user feedback or analytics? ChatGPT can help you analyze them, detect trends, possible groupings, then synthesize this data, and even visualize it, in table or other format.
Louis Blais, Product Manager at Thiga, recounts: "For my current client, I'm asked to analyze a file of several thousand lines and highlight major behavioral habits and insights. The task would have taken me long hours that I wanted to invest elsewhere. So I anonymized the file before submitting it to generative AI. In just half an hour, the time it took to refine the prompt, I had important information and, above all, graphs that could be presented to interested parties. In addition to saving time, the added value lies in the reusability of the prompt, which enables regular data monitoring in just a few minutes, without having to use a PowerBI type platform."
Automation level: 2/3
Type of requirement: discovery and quantitative analysis.
Draft a market study
If it's not advisable to rely 100% on artificial intelligence to carry out your next benchmark, ChatGPT can at least direct your research by giving you an initial overview of the competitive panorama.
Level of automation: 1/3
Type of requirement: discovery and ideation.
Prepare user interviews
No need to spend hours devising the questions for your next interview: let ChatGPT provide you with an interview guide structure. All that's left for you to do is reread, refine and add the most specific questions, which the AI wouldn't have thought of.
Automation level: 2/3
Type of requirement: discovery and copywriting.
Transcribing user interviews
You can now give your full attention to the users you interview: generative AI can take care of transcribing the content of your discussions in writing. Make sure, however, that you adhere to good practice in terms of consent and data confidentiality.
Automation level: 2/3
Type of need: discovery and writing.
Formulating the problem statement
Your user research has provided you with plenty of insights, but you need some external help to see through them and formulate the most relevant problem statement? Why not ask ChatGPT to suggest it for you?
Automation level: 1/3
Type of requirement: discovery and copywriting.
Finding new ideas
As its name suggests, generative artificial intelligence isn't just for performing repetitive tasks: it can also help you to find new ideas, either by starting from a blank page (ask it, for example, how you could solve such and such a problem reported by your users), or by challenging what already exists (ask it, for example, what it would improve on such and such a product, or how you could optimize your roadmap).
To go further, discover our article on how to use GenAI and ChatGPT as a Product Designer
Kevin Si enlightens us with an example: "I often use ChatGPT to generate new ideas. Recently, for example, we had to put together the roadmap for the second half of the year. I already had a few topics in mind, but I still wanted to see what ChatGPT had up its sleeve. . So I explained the context, the product vision, recent difficulties and successes, competitors, the link to the web application, and so on. Then, as in a user interview, I asked him which topics he would prioritize for the semester, for what reasons and with which associated OKRs. This allowed me to explore other avenues and key outcomes I hadn't thought of."
Automation level: 2/3
Type of need: discovery and ideation.
Editing content
Finally, don't overlook ChatGPT's primary purpose: text generation! ChatGPT can help you to edit certain mass emails, training materials, technical documentation... This will free up time that you can allocate to higher value-added tasks.
That's exactly what Kevin Si did when he took charge of a new team: "On this occasion, one of our customers asked us for a release note between 9 versions of a product. This represented over 18 months' work! As the former Product Manager was no longer with the company, I went to Github to gather all the developers' commits into a text file, then asked ChatGPT to synthesize everything for me. 5 minutes later, the release note was sent!"
Automation level: 2/3
Type of requirement: discovery, delivery and copywriting.
Mapping your product's UX
UX mapping techniques are as useful in the discovery phase as they are time-consuming... Just what you need, ChatGPT can buy your "user journey/experience/empathy maps" and other blueprints. Don't expect a perfect result, but rather a working base that you'll need to revise, flesh out, confront with your user feedback...
Automation level: 1/3
Type of need: discovery and ideation.
Doing an impact study
Are you planning to develop a new feature, but finding it hard to apprehend the potential consequences, side effects, etc.? Why not ask ChatGPT to shed some light on the subject?
Automation level: 1/3
Type of need: discovery and ideation.
Identifying relevant KPIs
ChatGPT can help you identify key indicators you might not have thought of or, on the contrary, narrow down your list of KPIs to those that are truly essential.
Automation level: 2/3
Type of need: discovery, delivery, ideation and quantitative analysis.
How to create a good ChatGPT prompt for Product Manager.
Standing instructions
Save time for your next prompts: give ChatGPT standing instructions that it will apply by default to all your future conversations. To do this, click on your profile in the top right-hand corner, then on "Customize ChatGPT". Note that this feature is available in the free version!
"Custom GPTs"
If you have access to the paid version of ChatGPT, you can supplement your standing instructions, which apply to all conversations, with instructions and context specific to this or that conversation, by creating "custom GPTs". To do this, click on "Explore GPTs" then "Create" and let yourself be guided to give ChatGPT all the context you want.
Thiga consultant Maël Guilluy is a master at using custom GPTs: "My custom GPTs are like extensions of my brain, specialized in different fields. So I can create custom wizards with not only specific instructions, but also a knowledge base fed by my own documents and data."
The art of prompting
Any good prompt must include at least the following elements.
- The context: Explain to the model in what framework your prompt fits. For example: "You're a Product Manager for an online video platform and you're about to conduct a user interview." Beware, if the AI needs to take into account elements of a previous conversation, don't forget to specify it, or even to remind it succinctly if they go back further than the "memory" of your chat.
- The task to be performed: This is the heart of your prompt: the action you're asking the artificial intelligence to perform. Let's go back to our example of a user interview guide: "List me the questions to ask, which should in particular ask users about their expectations and friction points generated by using our platform."
- The type of response expected: Tone, length, style or even format: don't forget to specify to ChatGPT the type of response you expect from it. For example: "I want concise questions, without technical terms, written in a formal style and divided into headings. Maximum 30 questions in total."
- Test your prompts: Chances are, your first prompt isn't the best. In this case, don't hesitate to iterate by specifying your request further. If, for example, ChatGPT doesn't seem to have fully taken on board the context given above, we can spell it out in more detail: "I'm a Product Manager working at [Company name] in the[industry sector, e.g. "digital health"]. Our target audience is mainly [type of audience, ex: "healthcare professionals"]. My role is to [specific responsibilities, ex: "develop new features for our online platform"]."
The best ChatGPT prompts for Product Manager
Let's take the ChatGPT for Product Manager use cases listed above and see how we could prompt them.
Writing Users Stories"As Product Manager for a movie streaming platform, write the User Story corresponding to the functionality that allows you to view a movie trailer. Your User Story should follow the structure "As... I want... in order to...". You'll then add the acceptance criteria."
Analyzing and synthesizing data
[Pre-load data] "The file I've just uploaded contains insights from user interviews. What are the 5 most recurring themes in these insights? And for each of these 5 themes, how many insights are involved?"
Louis Blais, Product Manager at Thiga, confides: "A prompt I use every week enables me to bring out usage data for a feature of my product in just three clicks, despite the fact that the file comprises several thousand rows with some fifteen columns. Here's the prompt: 'Analyze this file and highlight the behavioral patterns of the past week in the composition of each line. Highlight the most important sectors of activity over the past week, as well as the average number of items in the columns [X] & [Y] (specific to my product). Put the number of uses for the week in comparison with previous weeks in graph form'."
Using the same "conversation" each week, simply upload the file asking for the same work as previous times. It's also possible to ask if there are any important changes to notify compared to previous weeks.
Drafting a market study
"As Product Manager, what are the main trends I should know about the fitness app market in France?"
"What are the main Product Management consultancies in France, with their respective sales figures and the year concerned?"
Preparing user interviews
"As Product Manager for [specify product], you're about to interview a user. What questions would you ask him?"
Transcribing user interviews
[Load the recording beforehand, for which we will have imperatively collected the user's authorization.] "Retranscribe the attached recording in its entirety."
We could also add "Now, summarize it in 10 lines." or "What are the 5 main lessons learned from this interview?".
Formulating the problem statement
Here we can either load prior data, or ask ChatGPT to "assume" the problem statement. For example: "What would a Netflix user's problem statement be? Write it following the Job-To-Be-Done format: 1. Context: "when I..."; 2. Need: "I want..."; 3. Impact: "in order to..."."
Finding new ideas
"As Product Manager, what improvements could you make to Netflix?"
Writing content
"In order to write a tutorial for using Netflix, describe the successive steps to reset one's password."
Mapping your product's UX
"Creates a detailed "customer journey map" for a streaming platform user. Identifies each stage of the journey, associated user actions, feelings at each stage, potential obstacles and opportunities for improvement."
Conducting an impact study
"As a Product Manager at Netflix, I'm considering developing a feature that allows users to launch the playback of a movie at random. What potential impacts, both positive and negative, do you think this new feature could have (on users, the platform, the business...)?"
Identifying relevant KPIs
"As a Product Manager at Netflix, I'm considering developing a feature that allows users to start playing a movie at random. What key indicators should I track to measure the success or failure of this new feature?"
Use of generative AI by Product Managers: what about ethics?
ChatGPT is obviously a formidable productivity gas pedal for Product Managers, but its use raises ethical questions. Precautions are therefore called for.
The main ethical issue with ChatGPT is data confidentiality. So remember to anonymize the data you submit to it.
Another ethical issue linked to the use of generative artificial intelligence: critical thinking. In other words, a good Product Manager cannot simply delegate 100% of his tasks to ChatGPT. He or she will have to re-read the answers produced by the artificial intelligence, pushing it to refine them without ever forgetting to take them with a grain of salt.
In conclusion, generative artificial intelligence, and ChatGPT in particular, is a very powerful tool, enabling Product Managers to process certain repetitive tasks more quickly, to reclaim time for higher value-added tasks, or to stimulate reflection and creativity. It is therefore crucial for any Product Manager to be trained in the use of generative AI, in best practices as well as in the mistakes to avoid.
For more Product Management tools: download our Product Design Toolkit
