The increasingly common use of artificial intelligence (AI) is lightening the work burden of product managers (PMs), automating some of the manual, labor-intensive tasks that seem to correspond to a bygone age, such as analyzing data, conducting user research, processing feedback, maintaining accurate documentation, and managing tasks. (And we think of a PM as a kind of AI already, for the way she seems to keep the many moving parts of a product in sync!) According to IBM research, 21% of product managers worldwide leverage AI tools every single day.
Vladimir Antonov, the Head of the Product Unit at Avito, has noticed a very strong interest in artificial intelligence-driven tools that are built to filter out a product manager’s responsibilities. He shared this overview of various AI products, with a spotlight on the features that the smartest minds in the field find most impactful. With those features in mind, I offer a warning: Be very careful about using these tools with your company’s data.
Important disclaimer
Many of the profiled tools use AI for a number of different jobs. They might gather data to recommend ideas, evaluate suggestions and produce finished documentation. For the most part, this spotlight focuses on just a few key functions inside each product. Keep in mind, too, that your mileage may vary in accessing these tools. Some are available only in certain regions. And as with AI tools across the board, experts who deal in e-confidentiality recommend using in-house products.
1. Ideas generation
The ChatGPT plugin allows you to have different configurations. You can set it up to work with your current system and tools. Whether public, private, or alongside wayfinding helpers, chatting with visitors is easy, secure, and private. From the collected data, you can visualize it in a variety of ways, but with privacy and security emphasized. So far, so good, and that’s just a simple description of the straightforward way the plugin works. But is there any downside?
ChatGPT is one of the widely utilized tools for copy, marketing material, email, and document generation. In some contexts, people use it as a writing assistant, while in other contexts, it appears under the label of AI-driven content creation. Regardless of the context in which you use it, it’s important to recognize that this is a tool with immense potential that is powerful in both the mundane and in more high-level creative writing tasks.
In order for this tool to reach its full potential, product managers supply it with copious amounts of detail in user requests. They spell out (1) the intended audience, (2) the intended goal, and (3) the intended format of the expected output. A typical request might be aimed at producing something like a very focused, narrow prompt.
Picture yourself in the role of product manager. The audience is a particular cohort. Propose three methods to achieve a certain objective. The outcome should materialize in a specified fashion.
The better the request, the better the result from AI. With a well-crafted request, ChatGPT may reveal unexpected or innovative solutions.
2. Meeting minutes
At Avito, teams conduct video conferences using Kontur Talk, which features a built-in transcriber that converts meeting recordings into text. For added functionality, Avito employs an internal AI bot in Mattermost, called Neuroslav, that takes a transcribed audio file and creates a summary of the meeting or action items from it.
For outside use, a service such as Teamlogs also offers transcription, speaker separation, and in-browser text editing prior to download. Its integrated AI can generate meeting summaries, pull out main points, respond to inquiries, and find specific quotes in the text. These features cut down the time required for a person to go through extended audio or video recordings to find the information they need.
3. User data analysis
Chattermill is made for apps with tons of users, like BlaBlaCar and Uber. It uses artificial intelligence to do something that old-school analytics can’t: make sense of all the different ways people express their thoughts and feelings about your app or service. Chattermill gets to the heart of the matter, what’s really working and what’s not.
Rebecca Guitton, Head of Research at BlaBlaCar, shared that after adding a new feature, Chattermill waved a red flag about user understanding. So, the BlaBlaCar team took its findings to heart; redesigned the feature’s interface; put it through some rigged usability tests (we’re sure it aced those); and came up with some better and clearer explanations for users. What does this have to do with AI? It has to do with how AI does seem to be getting pretty darn good at being both expert and analyst for product managers.
Although Chattermill is capable of being scaled and is potent in processing language, it can be a bit of a time sink to set up and somewhat fussy to manage.
4. Forecasting
Amplitude AI, which is associated with the well-known Amplitude analytics platform, collects user data from a variety of sources and employs it to act in three main ways: predict behavior, monitor anomalies, and create on-demand visualizations.
Product managers can directly interrogate the platform with simple queries (e.g., What is the current conversion rate?), and it responds with visualizations of the answers, usually of the chart or graph variety.
Amplitude AI analyzes user patterns, generates hypotheses for the future, and provides suggestions for documentation topics. This level of automation can save companies an incalculable number of hours and resources, letting them direct more of their attention to product strategy. Amplitude also has a free plan and a $49/month plan for small businesses. However, using it well requires time for onboarding and initial setup.
5. Hypothesis generation for A/B Tests
Zeda.io is a product management platform that can automatically gather user data (or accept user data that has been entered manually) and suggest ways to improve a product. A frequently mentioned example has to do with using AI to make suggestions that would improve a product’s accessibility, for instance, by using AI to figure out that something like a larger font would work better for a certain audience (one that’s older, for instance) than the font that the audience is currently using.
Zeda not only generates roadmaps and prioritizes tasks but also can automatically create and test product hypotheses. This is especially useful for product managers like me who work with multiple teams or on complex products. As valuable as its features are, however, that $499/month cost is a big hurdle for many product teams, and its tendency to require workflow changes can also be off-putting. Zeda is also somewhat limited for use by product teams that don’t work in English.
6. Correspondence management
Missive is a program that combines task management with a unified inbox for team correspondence across multiple channels. Its integration with OpenAI can suggest prompts, translate text, or even refine the tone of a message, allowing for faster and more consistent responses from teams. Missive is often recommended for streamlining communication across the many different parts of a cross-functional team. While Thunderbird also serves this purpose, many users prefer Missive, though it does lack a free plan and has a somewhat limited number of integrations.
7. Reporting
An internal tool named Datapolk helps with data analytics and text and transcript management at Avito. It typically is used to create terms of reference (TORs) that accompany the research and reports we produce. An external tool that can also be used is Google’s Gemini, which, in its experimental release, can generate first drafts of the reports, documents, and project plans we need. Like ChatGPT, Bard can help us condense large data sets into a few focused insights; its performance, however, depends a lot on how well we specify what we want.
8. Documentation
ChatPRD is a specialized version of ChatGPT built for product managers. It features document templates that streamline routine documentation. The service claims it can cut the time needed to create documentation by a factor of four. ChatPRD allows users to refine existing documents using AI commands. Although there is no free plan, experts suggest trying it to see how it compares to general-purpose AI like ChatGPT.
9. Data visualization
For a long time, Tableau led the industry in data visualization. But services like it have grown in popularity, and even some built-in services from spreadsheet applications have greatly improved over time.
Take Tomat.ai, for example. This service works with equations and data in spreadsheet form. If you asked it to do what Tableau does, it might struggle. But it can do what the best visualization tools do: provide conclusions, clean data, or highlight key information. Another option is Napkin.ai, which is free and serves as a kind of text-to-diagram or -infographic service. Try to picture the difference between these two tools, you might imagine doing so by drawing a diagram or by visualizing one in your head.
10. Checking for errors in analytics
Another analytics platform, Julius.ai, processes raw data to find the trends and deliver them in visually comprehensible forms. Users upload the data they’ve collected, and Julius.ai suggests charts and graphs that present the data in a way that offers up some conclusions. Those conclusions can then be compared against the existing in-house analytics and serve as a nice way to double-check that the business is looking at its data in the right way.
11. Writing SQL queries with SQL copilot
Multiple Copilot solutions currently aid in the composition of SQL queries. Take, for instance, Microsoft Fabric, which is piloting a version of Copilot that can do such things as match the words in a query with the correct segments needed to complete that query, fix any coding errors that are part of the query, and answer certain types of questions that about half the people using a database would like to ask.
Expert takeaway
Artificial intelligence (AI) is rapidly evolving and finding broader applications in product management. Many of the tools described in this report, once just visions of the future, are now part of the daily toolkit for product managers. As they become even more user-friendly and the functions they serve become even more cross-functional, we forecast that the adoption of these tools will accelerate dramatically over the next five years.
Relieving product managers of more mundane tasks and decision-making where they should be focused on the strategic initiatives that drive innovation and growth within their organizations, AI in product management seems to be finding its way into the daily toolkit for many.
Featured image credit: Igor Omilaev/Unsplash
