Understanding the price of privacy

With ChatGPT, we can provide our users with a built-in team of five

We’ve spent the last ~15 years digitizing our products:

• We slogged through company-wide digital transformations to migrate data and turn manual processes to digital.
• We’ve climbed mountains of mobile design and development to satisfy users’ needs in a bigger way.
• And in the latest stretch of our journey, we’ve sanded our digital product experiences to minimize friction and delight users as much as possible.

Then, without warning this winter, a giant curtain dropped and revealed the next chapter of our products: it’s showtime for AI, baby.

It’s not that AI came out of nowhere; however, it was starting to feel like we were stuck on a plane that would never take off. But look at us now — flying down the runway!

There’s something different about this moment compared to other recent tech “fake-outs” like voice assistants, the metaverse, and Web3. They all promised to be game-changers, and I admit, I believed them. 🙃

• I walked around CES in 2019 and remarked, “voice assistants would transform UI!” — while not recognizing their relatively limited capabilities.
• I took a client out to dinner in 2021 and encouraged her to buckle up for the metaverse — completely ignoring the adoption challenge that lay ahead.
• I bought into the hype of Web3 and the Blockchain — even though I could never find a use case that could clearly communicate the decentralized value to non-tech friends.

Unlike those charlatans of the past, large language models (and ChatGPT in particular) have become the next big thing because they have awed us with their power. It feels like we have spent the past few months with our collective mouths agape, marveling at what ChatGPT can do.

AI is the next chapter for us, not because Zuck told us so, but because it has proven to be immediately useful, seamlessly integrating into our daily lives.

• The last sentence you read was clunky until ChatGPT helped me smooth it out.
• At a recent dinner party, my non-tech friend entertained us with a dramatic, martini-enhanced reading of ChatGPT’s legal writing.
• ChatGPT is breaking down language barriers for non-native speakers, allowing them to research and write confidently in a new language.
• Patients are turning to ChatGPT to quickly analyze their lab data and receive conclusions before their busy doctors can respond.
• Individuals who feel marginalized by society are finding comfort in ChatGPT’s non-judgmental daily affirmations.

The proof is in the pudding: ChatGPT’s has broken records with all-time fastest adoption rates.

And a surprising amount of big companies have already launched their ChatGPT partnerships/experiments/plugins.

Lastly, the economic implications of ChatGPT are already being felt. Just look at Chegg’s recent announcement that ChatGPT is slowing its ability to grow — cutting its stock in half.

Put simply, the transformers do be transforming.

Exciting, uncertain, and scary — all at once

I have to acknowledge, amongst all the hype, as a product designer, this is a weird moment. I’ve never felt so excited, unclear, and scared about technology — all at once.

• Exciting! The ChatGPT API is democratizing AI and opening up new possibilities for product design.
• Uncertain! As we embrace AI and large language models like ChatGPT, we enter uncharted territory where the long-term implications are impossible to predict. We don’t have the right metaphors, experience, or empirical evidence to analyze our situation.
• Scary! We learned through the social media era that we can’t trust Big Tech to prioritize our collective well-being over their profits. And the stakes are higher this time: half of AI scientists believe there is at least a 10% chance that this technology could result in humanity’s annihilation.

I must that this article is a cop out. I’m putting blinders on and focusing on immediate and tangible ways ChatGPT can enhance products.It’s overly pragmatic at a time when we need to focus on the bigger picture and potential consequences.

However! The first step in human-centered AI is to lead with people’s needs — not the technology — so let’s start here.

In the coming weeks, I promise to take the blinders off. I’d love for you to follow along on the journey.

The five member team

ChatGPT is far from perfect. It hallucinates, struggles with complex tasks, and is painfully literal. But just like an intern or junior team member, with a little guidance and patience, it can take on five basic tasks.

Think of your product as having a built-in team of five helpers. (This metaphor is inspired by Ezra Klein’s recent AMA podcast.)

Allow me to introduce you:

I love working with skilled analysts. They dive into the data — unthwarted by messiness or complexity — and craft a perfectly clean picture on the other side.I’m not just talking about business here. It could be anyone who has a talent for simplifying vast complexity into clear outputs. For instance, that friend who always knows the best restaurants in town and can recommend the perfect three for your anniversary dinner.

I love working with an analyst because tromping through messy data is overwhelming! And, frankly, I much prefer what comes after the data has been distilled and organized.

As an analyst, ChatGPT, is really good at presenting clear nuggets from mounds of messy data. You wouldn’t want today’s ChatGPT to make final conclusions or decisions on its own, but it instantaneously provides you the foundation to take the next step.

• The need/job to be done: the user is facing a decision-making challenge that requires a foundational understanding of a topic. However, the corpus of information available may be overwhelmingly large, inaccessible, or cumbersome to study.
• Use case: ChatGPT can analyze large sets of data and provide summarized insights and recommendations.
• Examples
• Decide what to make for dinner by inputting the contents of your refrigerator and relying on its analysis of potential ingredients/recipes.
• Expedia: plan a trip by relying on ChatGPT’s analysis of the hotels, activities, etc, of your destination.
• EPIC: make changes that improve operational efficiency by relying on ChatGPT’s analysis of EMR data.
• Stripe: verify customers faster by relying on ChatGPT’s analysis of what the company is/what they do.
• DataRobot: draw insights in natural language by relying on ChatGPT’s analysis of complicated quantitative data.

Life can feel like an unending crawl through logistical mud — whether you’re managing a family’s calendar, your project at work, or household inventory. Administrative tasks may not be glamorous, but there’s no way around them.

On top of that, the last decade of technological improvements have squeezed our attention spans like little raisins, which can make every day feel noisy and distracted.

Like a trusty administrative assistant, ChatGPT can step in to help minimize your user’s logistical efforts.

• The need/job to be done: your users often face mundane and repetitive tasks, like tracking statuses, sending reminders, or updating aggregating updates.
• Use case: ChatGPT will automate repetitive administrative tasks for your users.
• Examples
• Track a house renovation project by quickly typing the numbers into ChatGPT
• Milo: manage your household by sending Milo freeform dates, voice memos, and lists. Milo organizes it into calendar invites, daily rundowns, and reminders.
• Salesforce: Einstein uses AI to deliver instant conversation summaries, so that you don’t have to do this after a meeting with a customer.

I struggled with group learning in school. The fear of looking stupid prevented me from asking questions and many generic teaching approaches did jive with my brain. However, one-on-one tutoring, with real-time feedback and individualized support, unlocked complex topics like chemistry and physics.

Like a patient teacher after class, ChatGPT, the tutor, can provide real-time feedback and one-on-one support that makes learning easier. Unlike the analyst, the tutor is not doing the learning for you so you can make a decision; it’s supporting you through the process of learning.

• The need/job to be done: the user is trying to learn something new, but they’re doing it alone and need personalized guidance.
• Use case: one-on-one guidance throughout the learning journey
• Examples
• Ask ChatGPT how to improve negotiation skills when facing a rent increase, and receive personalized feedback and tips.
• Duolingo: Learn a new language with real-time feedback on what you’re doing right or wrong. Or role-play chat with a native speaker.
• Khan Academy: Ask your AI tutor questions to get help with a question.

The creative process is a strange mix of magical vision and mundane execution. The brain soars in the idealistic clouds, imagining what could be. Then it has to abruptly descend to the ground to grind that vision to life. This whiplash kills momentum and motivation. And many creative people are not good at both.

ChatGPT can act as an “apprentice” to assist with some of the more mundane tasks involved in the creative process — such as writing, designing, or coding. This apprenticeship model is similar to the way a master craftsman would train an apprentice in the Middle Ages, where the apprentice would take on much of the grunt work of executing the master’s vision.

This capability relies on some of the broader capabilities of generative AI; however, it allows us to bring our ideas for video games, websites, interior designs, D&D encounter, or novels to life, without having to delve into the technical details ourselves.

• The need/job to be done: the user is creating/designing/writing something. They have the vision, but executing a vision is time-consuming and requires technical expertise.
• Use case: ChatGPT can serve as an apprentice to execute the more mundane tasks that go into writing, designing, coding, or other creative pursuits. With clear direction and guidance from the user, ChatGPT can take on some of the heavy lifting and help bring the vision to life.
• Examples
• Use ChatGPT to edit your resume to match a job description
• Genius (by diagram): autocomplete UI designs in Figma
• Github copilot: autocomplete code
• Tome.AI: create fully designed decks like a graphic design apprentice
• Interior AI: create interior design mock-ups
• Google Workspace: write emails with customized tone and voice.

Think about an experience you’ve had with a great coach — whether that’s a therapist, personal trainer, or nutritionist. A coach is most valuable in situations where you want to achieve something but have no idea how to do it.

Instead of flailing about alone, without a foundational understanding of what to do, a coach uses their expertise to guide you on the path towards success, providing a steady hand along the way.

While ChatGPT may not be able to win the ‘coach of the year’ award, it is capable of helping to create a plan that can set the user on the path to success, especially when they don’t have access to an expensive coach or someone with an understanding of a nuanced topic.

• The need/job to be done: the user is striving to achieve a new goal or make a change in their life, but they lack the knowledge or direction to get started.
• Use case: ChatGPT can create a personalized plan based on the user’s constraints and help guide them along the path towards achieving their goal.
• Examples
• Get a workout plan to get bulked before the summer…that won’t exacerbate back problems.
• Get a plan to turn a wild puppy into a good boy in the first 5 weeks of his life.
• Get a plan to get out of debt when the holidays are coming up

This article misses the bigger picture

This article started with a functional, short-term perspective on ChatGPT, which feels safe and unencumbered by second/third-order consequences. But AI’s potential doesn’t stop there.

What happens when AI becomes our go-to companion, therapist, or favorite artist? What happens when AI use cases touch our deepest needs and the foundations of our human experience?

This is what makes me most uncomfortable. If we’ve learned anything from the last decade of social media, it’s that the products we create shape the way we live, and big tech companies don’t always prioritize our collective well-being over competitive pressure.

As product people, it’s up to us to figure out how to design human-centered AI.

Follow along as I explore this topic in this exciting, uncertain, and scary moment.

Works cited

• Ezra Klein’s AMA Podcast: https://www.nytimes.com/2023/04/07/podcasts/ezra-klein-show-ama-april2023-transcript.html
• AI Impacts Expert Survey: https://aiimpacts.org/2022-expert-survey-on-progress-in-ai/
• Humane Tech: https://medium.com/r/?url=https%3A%2F%2Fwww.humanetech.com%2Finsights%2Fthe-global-effects-of-extractive-technology-video

---

5 use cases for human-centered AI was originally published in UX Collective on Medium, where people are continuing the conversation by highlighting and responding to this story.