My "AI Product Management" utopia.

I have worked with AI products since 2017, and when GitHub launched AI Copilot, I wondered when there would be a copilot for Product Managers. Back then, AI and LLMs were rare, so I thought of building a product, making it a platform, and then making it open-sourced, getting data, training AI, and launching AI copilot for Product Managers.

But GenAI changed the game. While there are many AI product management tools and a plethora of GPTs, there is a long way to go, and hopefully, we at Zeda.io will lead the way.

Here are a few use cases where I believe AI will shine in Product Management -

1. Figure out which problem to solve based on goals, user needs, and market.- One of the biggest questions is - "What to build?" that will help achieve business goals. I have talked to countless product leaders, and while everyone wants to innovate, they all are risk-averse because of costly failures. As a PM, there are always items in the backlog to ship, sales teams and users are always asking for something, and there is a new product or change in the market now and then. So the big question becomes what shall we build to help us achieve business goals and user needs based on our current talent or investment budget within [X] margin of error. We know so many products that are shipped, and no one wants them. AI can help reduce this wastage and ensure companies invest in the right products.
2. User research and validation - Once you have identified the problem to pursue, the AI should also present you with everything in your company database and in the market that will help you build conviction and direction. AI can go through your company information to highlight historical context, customer communications, product areas that will be impacted, and people who need to be consulted. AI will also monitor competition news, scan the Internet for relevant articles, and help with market sizing and any other information needed for business cases.Then, AI can provide a list of users who can be reached for research and create a draft message. AI will also help with recording and transcribing these sessions and giving insights. AI will also help brainstorm by playing devil's advocate, an expert, or a user persona. Lastly, taking all this information and presenting it in a useful format to be consumed by other stakeholders.
3. Actual useful PRDsLet's be honest: Most of the PRDs currently generated by AI models are generic. They don't consider the context of your user, business, product history, technical architecture, etc. Ideally, AI can understand your product wiki, research, business requirements, and current product and come up with a good first draft that you can edit. Even for generic use cases like sign-up flow, account page, notification center, etc, the current quality of PRDs created is not helpful for high-stakes products. So, I expect this to change soon. We have already seen huge strides in design, development, sales, marketing, and support for PMs.
4. Improvise your requirementsI know that if you are building a login flow, you need to have a forgotten password, you need to resend OTP there, and you need a rate limit on that API. I have messed up. But now, for a junior PM to know not to mess up, the only way is either to make this mistake or for me to review everything they have written. Both are costly. AI can help here by taking these scenarios, filling in these requirements, and improving the product. Another example: Alibaba has researched an app for elderly people. Now, if someone is launching a product for elders, AI should automatically suggest making the font and icon bigger. Similarly, if for kids, automatically apply nudity and safety filters; if for EU, apply the GDPR checklist.AI can improve products from their inception. We all know the poor experience with airline apps, government apps, and banking apps. AI can make all these products default good, not great, but without glaring issues.
5. Make prototype and user flows- Once you have identified what to build, researched, and defined the product. AI can help create a user flow and prototype to visualize and share with stakeholders to get thoughts. I have spent a lot of time on Balsamiq, and I don't want to spend more.
6. Prioritize roadmap and resources.- This is the most annoying thing in bigger companies with dependencies across different teams.Imagine a company with the goal of accepting Bitcoin payments and launching the Vision Pro app. You will have a design team, accounts team, payment team, app team, backend team, and one or more teams involved in shipping these two requirements. These teams will assign tasks with different priorities based on who has better relations, power, say, etc. We will have to hold multiple meetings, call team leads, spend time, and build alignment. Wouldn't it be great if AI reviewed everyone's backlog and resource availability and prioritized tasks optimally, just like a math problem? It should always be user first, company second, team third, and individual last. AI can help build alignment across different teams or at least make a proposal so people have a hard time justifying their egos.
7. Streamline product information- Over the last four years, I have heard multiple times that in bigger companies, people don't know whom to talk to for feedback or complaints. People come and go, and with them, the product knowledge. Documents are out of date. GTM teams don't know all the features of the product. Support teams get the exact tickets and resolutions. The product information is spread across various tools, and not everyone has the right access. By the time the information reaches leadership or the right person, there is too much personal bias, or it is too late to take any action.AI can help with all of this. This is a nightmare. This was the starting idea of Zeda.io : to be a place to visualize the product's past, present, and future.
8. Streamline communication -Product managers are an expensive resource, wasting time in useless meetings that could have been emails or Slack messages. There are so many changes to the product, the roadmap, or a decision taken without all stakeholders knowing it.An AI should be able to ensure everything is up to date and available to all stakeholders at their fingertips. Not to forget, writing is crucial for product managers. No wonder most companies use documentation or Excel sheets to run their entire product management division (Jira is for dev teams, not you). AI can help improve writing. This use case is currently taken care of by AI, but I am adding it to the list because it came across my mind.
9. AI Oracle - This is a science fiction vision, but hear me out. AI will run simulations on all permutation combinations of all data points to tell me the probability of success for my product strategy.Run a test to see what would happen if you had prioritized A vs. B, allocated X vs. Y story points, executed various GTM strategies with different types of users, and so on, and finally came up with a path with the maximum outcome and least cost.If you have seen Black Mirror's Hang the DJ episode or the Devs series, you'll have an idea.
10. Make executive reports and decks - This ask has come up at least five times during my sales calls in the last two months. Clearly, PMs prefer to avoid creating presentations for managers, who will then use them for their leadership all the way to the boardroom. This practice from our "MBAs with engineer background" days will haunt us for a few years.
11. Provide the best framework or method for a problem - This is all thanks to thought leaders. Everyone has a framework, best practices, and gyan. It's good to read and makes sense. But I forget. When the actual problem comes, I don't recall frameworks. So, if an AI copilot can suggest frameworks and best practices at the right time in the right place, I will pay for it.
12. Help with analyticsI am not a big numbers guy. I hate Google Analytics. How is it performing, what could be improved, what went wrong? I still hate all the GA events to be implemented with every release, managing different reports, and going through a haystack to find a needle. Why can't AI summarize my takeaways and highlight what needs attention? SQLs are already sorted, and tools like Segwise.ai and June.so are trying to make this possible.
13. Post launch analysis and retrospectives -This can be in two parts. Quantitative and qualitative. AI should tell me how much the product achieved my OKR or KPIs. DoubleLoop is doing a good job. But still, it doesn't tie back to the performance reports, OKR tools, and, most importantly, my stakeholders. Everyone on the team should know how the launch is performing in an easy-to-understand manner.AI can also learn from this launch and process and ensure things are better next time.

This is the big picture and covers the main four pillars of product management: product discovery, product planning, product building, and product launch. We at Zeda.io are working on the first two and aim to cover the latter two in the future.

AI will help product managers claim back time from operational tasks so they can focus on the art and craft of product management. This will raise product quality by default. The future is bright, forcing product managers to move beyond all the frameworks and disguise being busy to actually shipping value for users and businesses.

What are your thoughts on future of Product Management with AI?

Let me know if I missed an interesting use case. I would love to hear.