How GenAI is Changing Product Development
AI is moving fast, and requires us to rethink everything we know about building product.
Product innovation isn't just about refining what we know, but about venturing into the unknowns. As we stand on the cusp of a new era, generative AI is more than just a new technology—it's changing our world quickly. $23B has been invested in USA AI startups in 2023 alone, representing 25% of all investments [2]. Startups are exploding with funding and users are willing to pay for the immense value that AI can deliver.
Companies who can establish AI solutions now, have the advantage of capturing early market share and becoming household names of AI products. It's time to build rapidly because AI is changing everything, and it demands that we also quickly adapt the way we approach designing product altogether.
Create your IP through a unique user experience layer.
When we look out at the world, it's easy to say "every AI experience should be a chat bot" because that's what is really taking off right now. But there's so many more possibilities to how generative AI can be displayed to and experienced by users. Understanding user needs allows a team to generate product ideas and narrow in on solutions.
In the world of completely open-ended chats, the onus of creating good questions to ask is in the hands of the user.
A big open message box is intimidating for users.
As designers and creators, we have the power to craft guided experiences on top of LLMs (large language models that fuel generative AI such as chatGPT). By understanding your users distinct needs, you can craft user experiences are differentiated in the AI market. Through thoughtful and guided experiences that serve specific purposes, product teams have the potential to create stickier experiences that actively drive value.
It's easier than ever to build with AI technology. I am astounded by the amount of API access. You have to think outside of the box in order to differentiate yourself. It's not enough to be a chatGPT replica, but a user experience layer is needed.
How do we create products people are willing to pay for?
TLDR; Design thinking is still needed for creating AI products that solve real human needs in order to prove product-market fit.
Since the launch of ChatGPT in the end of 2022, there has been an explosion of generative AI products. AI is completely new and moving quickly. There is a unprecedented amount of open source APIs and accessible resources to build, so finding the right problems to solve for has never been more urgent.
In order to find the right idea worth investing in, try evaluating features early to be sure that the idea has an audience and a business model. By finding solutions that solve real problems, you can often directly find an audience worth paying for them.
Finding your AI + business match
By discerning user needs, we can design unique experiences in the AI domain. I'm inspired by several innovative businesses I see today in the AI market. The most successful companies have been able to match AI tech capabilities with specific user experience layers that collectively create unique value for users.
Here are a few tips when adapting your Design Thinking process for an AI feature:
How can we leverage AI to solve a user need?
Let's take a dive into a case study of Duolingo's new tier, which has successfully leveraged AI to create new value that users are willing to pay higher subscription prices for.
Duolingo Max Case study: Creating a new subscription tier powered by generative AI
Duolingo's DuoMax subscription tier launched in Q1 of this year. The monetization team built this product in just a rapid 5 months. The new subscription tier offers an incredible amount of value to users. Learners love it, and Duolingo has seen a boost of 62% dau growth, 44% revenue growth, and stronger profitability in 2023 (3).
Duolingo has further differentiated themselves as not just another language app. By leveraging the power of AI, the could launch a dynamic a role-playing experience as well as deep personalized explanations of language errors that help learners improve more quickly. These new features plus all existing premium features are available to learners at the highest tier of $19.99/month. The value has proved to be worth that price point.
As someone who works on the monetization team at Headspace, I am so inspired by this story of creating this new subscription tier is a great example of how AI is changing product development.
Step 1: Opportunity
Starting broad with the possibilities of AI, the first step is to narrow in on how your company and AI can work together. For example, AI and Education make a good team.
After exploring an early demo of ChatGPT, Duolingo's CEO Luis von Ahn said to the monetization PM Edwin Bodge, "Pivot your team immediately-AI is going to transform education." (4)
The monetization team put everything else on the roadmap on hold to prioritize creating a solution utilizing this powerful language model technology quickly. They had a goal to develop an entire new subscription tier utilizing AI in 5 months.
How might Duolingo leverage AI for learners?
Knowing the power of generative AI for learners, the team started work backwards to narrow in on the specific problem to solve for learners with AI that fit their business brand identity.
Step 2: Empathy
It's up to your company to decide who you will serve. For Duolingo, it's learners. Duolingo started with the technology and then went broad, generating possible user needs.
领英推荐
Instead of narrowing in on one specific user need like in normal design thinking, the team started with a vision for the AI technology and went broad to generate possible need state matches.
Many teams already have bounds of knowledge and data on their users. AI challenges us to take a new microscope to past data, and revisit. Knowing the power of generative AI for learners, what user needs might we solve for? If your team is new and you don't have any research to draw upon, start talking to your users to generate possible user needs you can solve for.
These are the 5 possible needs that the team considered solving after their generative phase revisiting existing research.
This is a generative step that will be narrowed in on next.
Step 3: Define
Duolingo hosted an internal "prompt hack-a-thon"and brought together about 20 people from cross-functional parts of the company to try prompts. They sat in a conference room firing off to try to imagine what would work for their millions of learners. I love this approach from Duolingo's brilliant monetization team.
Prototyping can be as easy as typing in chatGPT, but you will want to prototype with the AI model you plan to use. This is key, because testing with the AI model will help you define what works. There are some things AI does well and others that it doesn't do well. As product creators in the AI space, it's our job to set guard rails and define the experience in a way that delivers value to customers.
Keep the previous steps in mind as you narrow in. For example, Duolingo is known for their dynamic characters and easy learning patterns. As they tested and iterated on which prompts work, they kept in mind the playful characters that define the Duolingo brand, exploring how they could train chatGPT with parameters to make moments that felt silly and memorable.
As you narrow in, ask yourself is AI capable of solving the known user need predictably and reliably?
As you try and solve the user needs with AI, you might find that some needs are a better fit to solve for than others.
By marrying technical possibility with the needs that are possible to be filled by that technology, Duolingo was able narrow in and invest in only the most valuable features through trial and error during design.
Step 4: Launch and Learn
AI, by nature, is dynamic and open-ended. It is responsive and powerful. It's not always possible to test AI experiences with a figma prototype. Learning from the live experience is more important than ever with AI features.
Launching is a a divergent process because as you have more beta testers, you will continue to learn and refine your solution. By actively be listening and observing how users engage, you will have a natural feedback cycle for both qualitative and quantitive data.
Then repeat it all. Go back to refining the technology capability, and generating possible ideas. A great team will observe results, summarize learnings, and generate ideas following first launch together in order to keep evolving a product.
Beyond a linear journey
The product team at Duolingo did the work of creating for the learners they serve. Learners may be able to reach similar outcomes from the open text field in ChatGPT, but Duolingo Max created a curated role playing and personalized language coaching for language learners. This is what building AI product is all about.
Generative AI is more than about solving a problem one time; it's a about creating dynamic experiences for users to spend time in. It's about giving users the tools to explore, create, and discover new parts of themselves. Instead of straight paths that we've created tech solutions for solving problems in the past, with AI we are looking at branching journeys that evolve with each user interaction to continuously deliver value.
Generative AI changing our world, and we need to evolve product development to adapt. AI enables us to go beyond solving problems with UIs, and think broadly about creative toolboxes and co-pilots that can lead to exponential value.
Instead of linear design solutions, we need to invest more energy into the branching long term journey maps that evolve as a user engages with a model.
In Summary
While chatbots might be the current rage, generative AI offers a plethora of user experiences. In this open-ended chat landscape, the onus is on us, the creators, to guide user interactions. We must know our users and curate AI product experiences that fit their needs. Design is pivotal in shaping experiences atop AI technology to deliver tailored value.
By discerning user needs, we can design unique experiences in the AI domain. Thoughtful, purpose-driven designs can foster long term user loyalty and value that's worth paying for. By starting with AI tech and following this framework to narrow in on a specific use case, you can carve out a niche in the booming AI market.
More coming...
This is part 1 in my 10 part series exploring building product for AI through case studies and product philosophy. Other questions I plan to explore: How do we create feedback cycles that respond to users behavior? How does a user experience with the AI model deliver value over time? What role does the human in the loop play? As product creators, it becomes our job to design the guard rails as we drive down dynamically evolving roads alongside our users.
Thank you for reading.