How GenAI Could Boost Product Development

A four-stage, AI framework makes the case for streamlining app-dev from concept to customer

By Atin Gupta and Geoffrey Parker

Imagine developing software products so quickly and cheaply, your company could test more ideas with real — even virtual — customers. As a result, you could bring targeted products to market much more quickly.

Generative AI could make this possible. How would AI-powered product development work? We propose a four-stage framework using GenAI to improve the process over today’s human-intensive practices.

To be sure, we’re still far from this approach. As Silicon Valley Product Group leader Marty Cagan has explained in his writing, most companies still develop new software products using a long, complicated process. First they come up with an idea. Next, they develop its business case and create a road map. Then they determine the product requirements, followed by designing, building and testing. Finally, at the end of this marathon, the company actually deploys the new software product.

Developing software products can also incur risk. Development is expensive, yet calculating a product’s ROI before release is typically incorrect. Also, getting customers to test product designs can be slow and expensive.

In addition, if the final product fails to attract customers, a great deal of time, human power and money are wasted.

(There are countless examples of failed products, with Quibi and Clubhouse being notable recent cases.)

Some firms have responded by building relationships with end users to better understand their needs. Developers then work with designers and engineers to quickly test prototype solutions that address those needs. This kind of A/B testing can lead to fast, incremental innovation, while also eliminating the need to “hit a bullseye” on the initial launch. While this approach addresses part of the product-development challenge, further change is needed. GenAI can drive that change.

The GenAI Transformation

GenAI has already shown product-development capabilities that transcend A/B testing.

For example, when asked whether his company does A/B testing, Nikita Bier, Product Growth Partner at Lightspeed Venture Partners, recently posted: “No, I just ship the app — and if it’s not ranked in the Apple Store, I change it until it is.”

That may seem bold. But with GenAI, that kind of boldness should be within everyone’s reach.

To help you see how we’ll get there, we propose a simplified, four-step framework for product development. We also have some ideas about how GenAI will help at each stage.

Stage 1: USER RESEARCH

How it’s done now: A company explores a user problem that needs to be solved, along with its context, all to answer the question: Why build this product in the first place?

How GenAI will help: Because the technology can simulate realistic consumer behavior, it could fill in for human customers. For example, in a recent study, researchers asked OpenAI’s GPT-3.5 platform whether it would buy various laptops at different prices. When GPT was prompted to understand that its annual income was $120,000 vs. $50,000, the platform became less sensitive to price, just as a human would.

Stage 2: DESIGN

How it’s done now: Product developers aim to create an effective solution to the user problem. They also try to determine the interaction between the product and its users.

How GenAI will help: The technology can be used by both visual thinkers and those more comfortable expressing ideas in words. Visual thinkers can sketch their designs on a digital tool, then use GenAI to convert it to formal design assets. Those more comfortable with words can use AI tools such as Galileo and Genius to convert natural language descriptions into wireframes. Then these designs can be imported into design tools such as Figma.

Stage 3: BUILD

How it’s done now: The development team determines how the product’s pieces fit together, then make it work with coding and other tasks.

How GenAI will help: Because software code is also text, GenAI tools can learn large amounts of code, then generate more of the same. The extent of this power can be surprising. For example, when aerospace structural engineer Brandon Starr wanted to instruct the code-writing program Replit Agent to create an entire app, he did so using just one sentence: “Create a bunny themed flappy bird as iOS app.”

Stage 4: LEARN

How it’s done now: The company determines whether the final product is solving the customer problem effectively.

How GenAI will help: The technology will be added as a top layer to best-in-class tools for product analysis. In this context, GenAI will synthesize data from these tools, then feed the findings back to the Design stage, enabling automated improvements, builds and relaunches. GenAI can already do much of this, as Ethan Mollick, a professor at the Wharton School, has demonstrated using GPT’s advanced data analysis feature.

You might wonder about market research, segmentation, feature prioritization and other traditional product-development steps. Some are now included in our four stages. Others — such as market analysis — will become less important as software development gets faster and cheaper.

More to Come

Yet another development — the introduction of a natural language interface — could streamline the four-step process even further.

This interface could guide a human product developer through the four-step process, regardless of how vague or specific their initial product idea might be. Then, the interface could not only develop the software product on demand, but also assess the product’s likely viability as a business.

If these AI tools are implemented, they would spur huge changes in how companies develop software products and organize their developers.

GenAI may also bring software development capabilities to individuals, much the way a previous set of technologies brought video creation — previously the domain of large TV studios — to anyone owning a smartphone.

At a time when GenAI is pioneering the way forward, software development may be one of its next frontlines.

About the Authors

Atin Gupta is VP of Strategy and Innovation at BuzzBoard.ai, a provider of Gen AI-powered sales and marketing services for small and midsized businesses.

Geoffrey Parker is a Professor of Engineering Innovation at Dartmouth College and a Research Fellow at the MIT Initiative on the Digital Economy (IDE).