Consumer AI is here — but people don’t know how to use it
When I first used ChatGPT in November of 2022, it felt magical. People always told me that when the IPhone was released it was also “magical”. Once every decade it seems, a piece of technology comes to market that is truly magical. It isn’t just a new feature or upgrade, it’s an entirely new technology that feels impossible it exists today. ChatGPT initially felt that way.
No doubt AI is amazing and has some incredible use cases in consumer products. The problem is, people don’t know how to use it.
————————————————————————————————————————
I have used AI in my day to day life for over a year now. Whether it helps me with emails, converts PDFs in excel files, or helps me write new content, it has been extremely helpful in what I do. I have also seen a few colleagues and friends take advantage of the technology. All to often I am hearing that AI and ChatGPT “isn’t good at certain tasks” or “didn’t really help me”. I have also spoken to my parents and their friends who say similar things like “I didn’t really get what I imagined out of it”.
Speaking generally about ChatGPT, the main issue is with prompts. With most people, they simply do not provide enough information or direction for AI to accomplish the task properly. I always say the general rule more information and direction = better results. See the chart below.
For myself, my prompts usually average over 100 words and reference my own tone of voice, information that is key to the task, and extensive directions. While it would be awesome to ask the AI to perform an action in a sentence, the reality is smaller prompts are too general and the AI cannot properly perform the task.
So TLDR; if you aren’t liking AI’s outputs, extend your prompts. Now to move on to AI in consumer applications.
Integrating AI into Consumer Products
Since the rise of popularity in AI, many companies rushed to put AI on their product roadmaps, or build entirely new AI products all together. A large majority of these were rushed out of the door and I am guessing not living up to the success metrics they were looking for. I would bet that most have fallen dramatically under. This isn’t because the feature or product wasn’t needed, it’s because consumers really don’t understand how to use AI yet. When launched, these products really need training wheels for most users to realize how they should use it. This either comes in the form of prompts, notifications, or just surfacing insights from the AI.
I’m going to highlight one company that I think is doing AI not right and one that I think is doing AI right.
Whoop → Needs Improvement
I’ll start this one off by saying I am a big fan of Whoop. In fact, I have had a Whoop on my wrist for over 1500 days, I don’t know what my life would be like without it. That being said, I have not been impressed with their rollout of AI.
Whoop launched Whoop Coach, powered by AI in September of 2023 and I am guessing at this moment, the usage is low. They marketed it as “a search engine for your body” and in reality, it is only really good at answering general questions with some of your specific data included. I recently asked it to tell me my average mile times on my runs when I get 70% of my sleep need. It wasn’t able to do this. Now, this touches more on the general performance of the AI, which isn’t precisely product-related, but I believe it extremely important when launching any AI-related feature. The AI needs to be great.
On the product side, the AI is surfaced right in the middle of the page and provides you with a few general prompts when you click into it (photo below). The prompts are general and in my mind more for newcomers to Whoop. In situations like these, I would stray away from general prompts and try to talk more about what you do know rather than what you can know. Use my data to surface these insights before I ask. Tell me more about my sleep performance and what other users have reported helping their sleep. Tell me how to get my average pace on runs down from 8:45 to 8:15. Tell me how long it would take me to train for a marathon. These are all things that I believe can be surfaced with the help of the AI. I can take this insights, dig in more, then ask more questions with the AI. In the current format, Whoop give users a starting point, but no real reason to stick around.
I also believe these insights can be surfaced outside of the AI itself, namely like notifications. Examples could range from anything like “It’s a high allergy season, and you lose 11% of your sleep during this season, prioritize extra rest tonight” or “You just ran 3 miles at a 8:10 pace, this is likely a result of your good rest”.
The first step is to improve the AI capabilities, after that push the insights forward. Once people understand what the AI can do, they will understand how to actually use it.
Rogo → Doing it right
A few notes to start off here. First, Rogo is an entirely AI-focused product and I personally just really like the way that they have built the product for their customer base. On that note, this customer base is very niche. They sell their product to hedge funds, investment banks, PE shops, and anyone who needs a financial analyst (which happens to be an AI).
For their end user, they are have built a hyper-specific solution. Imagine a financial analyst at a hedge fund. What do you imagine them doing? Building charts, reading 10-k’s, analyzing earnings transcripts, and doing this while working alongside the companies policies and internal data systems. This is exactly what Rogo does.
Their user base, the associates and mangers at these funds, are usually pretty tech savvy, but as the get older, this starts to fade away. And as these managers are the key decision makers, the AI has to be simple and easy to use. From my understanding, the product is not just a search bar with a file attachment. The product has many different was to analyze events, earning calls, and reflect on your personal firms data systems. If you want to ask questions about a companies recent earnings transcripts, you can easily do that. If you want to dig deeply into a thesis, all while using your firm’s proprietary knowledge, Rogo’s AI can do that too.
Final Note
For a product like ChatGPT, you can stay general because you are the google of AI. For companies like Rogo and Whoop you need to help your users get off the ground when it comes to usage. You need to lead your users with prompts, notifications, and insights that push them to understand how the AI works and use it more effectively and more often.