Mobile app development in the Product Discovery phase doesn’t yet involve writing code. It’s the initial stage of a project where we gain knowledge about users and the market, define business goals, develop app ideas, and test them.
Discover how AI can facilitate decision-making when selecting app features, streamline research, and accelerate prototyping.
AI supports mobile app development in the Discovery stage
The Product Discovery process takes place before developers start building the app. This is when we develop the app’s vision, determine what new features should be included in it, and also which ideas to abandon.
Contrary to the concerns of many, artificial intelligence will not do all of this for us. Creative thinking, empathy, and a deep understanding of the context play a significant role at this stage. It is still necessary to independently conduct interviews, draw your own conclusions from research analysis, and develop your ideas.
However, AI can support us in various ways. It can, among other things, serve as a critic, accelerate certain processes, or point out areas we might be overlooking but should focus on. When used appropriately, artificial intelligence can minimize the risks associated with building new mobile apps, effectively allowing us to:
- avoid costly mistakes, for example, those related to selecting a financial model or the functioning of new features
- find out which of our assumptions are not data-backed and are based only on speculation
- save the time that would normally be spent carrying out some tasks in a conventional way, which would also result in wasting money and increase the risk of oversights.
Stage #1: Research – gathering data and analysis
Market and user research play a key role in the Product Discovery process. Knowledge about the target audience allows us to reduce business risk and create solutions that address real problems. Conclusions drawn from the research also open the team up to new perspectives.
Desk research–how to find information faster with AI?
There are many methods for acquiring information about users and the market. One of them is desk research. As the name suggests, it involves searching for data that can be obtained without leaving your desk. This includes information from various studies available online, related to the area you are interested in. These may include reports from other companies or organizations (e.g., Pew Research Center) or scientific papers. Reports prepared by your own company, for example, from sentiment or behavioral analysis tools, are also useful.
If you want to obtain such information, reach for deep research tools.
Deep research mode
When data from reliable sources is needed, searching the internet independently can take ages, and many questions still remain unanswered.
The solution then might be the deep research mode available in many tools like Gemini, ChatGPT, Claude, or Perplexity.
Source: Gemini (https://www.gemini.google.com)
It allows for conducting in-depth analysis, based on a larger amount of data. The deep research approach enables a closer look at individual issues. The main topic is broken down into various sub-topics so that the resulting report contains information from different perspectives. The final document comprehensively describes the issue the user is asking about.
Tip
The result obtained in deep research mode depends, among other things, on how you phrase the prompt. You can use a regular Gemini or ChatGPT agent, which will prepare the appropriate prompt for deep research on the given topic.
Example
Analysis of the state of eCommerce industry in Europie
1. Prompt generated in Gemini
2. Generating report in Gemini Deep Research
Source: Zrzut ekranu Gemini
From my experience, this process is still time-consuming even with the help of AI. After all, information sources need to be checked. Even the best prompt will not guarantee that all reports and articles are based on reliable sources. The difference, however, is that I ultimately find many more credible data points. I would never have reached many of them without deep research, which is why this method is worth trying.
Tip
After checking the data sources, you can add them to the NotebookLLM tool to quickly verify which fragments of the source materials AI used to generate specific information.
Do you want to test other AI research solutions as well? Here are a few suggestions to check out: SciSpace, manus, Research Rabbit, scite_, Consensus, Elicit, Sider.
Interview analysis
You now have general market data. It is time to learn the perspective of individual people who make up your target audience. For this purpose, you should conduct interviews with them. Even 5 people from a given segment are enough to gather valuable information and find out what their needs, problems, or concerns are.
The interviews themselves must be conducted using traditional methods – nothing can replace direct contact with the respondent. However, you can use AI to analyze the collected information.
Source: Dovetail (https://docs.dovetail.com/help/data-summaries)
There are different ways to approach this. One of them is using the Dovetail platform (https://dovetail.com/). This tool offers many possibilities, for example, it analyzes recordings from your interviews, creates transcripts from them, and generates summaries. Depending on your needs, these can be traditional summaries, but there are more options.
For example, you can generate a summary by topics, which facilitates the identification of similar patterns indicating respondents’ needs or problems.
For usability tests, the function that summarizes the most important tasks along with a reference to specific test moments is useful.
Remember that you should not completely entrust the analysis to artificial intelligence. It is still only a supporting tool that can shorten your work time and direct your attention to issues that are easily overlooked when using a traditional approach.
Stage #2: Brainstorming
This is the most creative stage of the project. It is not worth relying solely on ideas generated by AI, as you will then lose the opportunity to introduce something truly unique.
So, how can you make brainstorming easier with the help of artificial intelligence?
Persona
A persona is the profile of a target user. This is the persona we keep in mind when working on specific solutions for the app. A persona represents the characteristics of the people who are supposed to use our product. It is best to create it based on research data.
Personas vary, but often include information such as name, occupation, lifestyle, needs, fears, preferences, current ways of coping with adversity, and motivations.
How do you go about preparing a persona with AI? Approach it in stages. Let the AI agent learn the context of your project and ask you questions about your persona. After each answer you provide, it should ask subsequent questions that will allow the persona to be created.
You can also add anonymized research data to a project created on a selected platform (e.g., Gemini, ChatGPT, Claude). The chat will create the persona itself based on the shared information. A persona created in this way is a great starting point for teamwork during the product discovery phase.
Always remember good security practices when working with artificial intelligence. Do not provide AI tools with sensitive data and always verify the answers it generates and the data sources it relies on.
Affinity mapping
Similar to thematic summaries, AI can be used to group similar ideas. Each group creates a separate category. This method is called affinity mapping.
This is a tedious task that normally takes a lot of time, but an AI Agent can handle it quickly.
You don’t even need to use any special tools. The regular agent in Gemini or Copilot that you usually use for work is sufficient.
Stage #3: Prototype for usability testing
You now have a detailed idea for the app, so it’s time to test how users will perceive it. It’s better to find this out before software developers spend many hours writing code, which is usually costly.
That is why it is worth conducting usability tests on a prototype. This will tell you whether a feature is intuitive and what representatives of your target audience think about it.
Importantly, your prototype does not need to have all the features and views that you want to include in the final app. It is enough to show the user the most important paths and key features you want to test.
How to create an app prototype with AI?
There are many tools that enable fast prototyping using AI. Popular solutions for mobile apps include Rork, Figma Make, or MagicPath. You just need to enter a prompt, and the app will create a mobile app design.
Source: Figma Make, prompt: mobile app for chefs
Such solutions are rarely suitable for production environment due to technical reasons, but they work well when we want to quickly check something, test it, or visualize our ideas.
It is worth describing what the app is for, who is supposed to use it, and what style interests you. You can even utilize the branding of your website, and the tool will generate a prototype that is visually consistent with it.
Product Discovery with AI: conclusion
Product Discovery is a key stage in the mobile app development process, which facilitates decision-making and limits the risk associated with software investment. Thanks to AI, Product Discovery can proceed more smoothly and efficiently.
Artificial intelligence will not replace humans when it comes to tasks requiring creativity. However, it can support them by verifying ideas or showing alternative directions or solutions.
Contact us if you want to create a mobile app and need the support of specialists who know how to conduct the Product Discovery process effectively.
