How I used (and didn’t use) AI in a Product Discovery!

By Ali Vahed




Can AI revolutionise product discovery? Here’s how I experienced it with a real problem for one of our clients!

What was the problem statement? 

Registration and conversion rates were declining for an online brokerage client. Fix it!

Were there any constraints? 

  • Marketing and lead generation were out of scope!
  • KYC requirements were non-negotiable!


How did I initially approach the discovery (no AI yet)? 

I kicked off the discovery with a full-funnel analysis using Snowflake and FullStory to map the user journey from acquisition to onboarding. I identified drop-off points and measured the performance of each step, especially the registration and KYC process. Not good! 

Product Insight: Registration and KYC are non-value-added steps for users. They're necessary for compliance but not for the user’s primary goal, in our client's case: Trading!

Every extra step delays them from their main "purpose"/ "job to be done".

What was my main product strategy?  

Break the registration flow into atomic components (aka user stories),  streamline the flow with four principles: 
  1. Remove unnecessary steps (e.g., eliminate redundant data collection, integrate with SSO solutions)  
  2. Simplify if possible (e.g., improve timing and UI/UX of email validation) 
  3. Defer if possible (e.g., move the application form post-login) 
  4. Retain only if it must stay (e.g., keep KYC as is but educate users and delay it until necessary)

What was the outcome of the discovery?

We defined a lean product roadmap and an MVP structured around four clear steps: 
  1. Sign Up: Basic access (email, Password, and T&C) 
  2. Application: Collect compliance essentials (Name, DOB, Contact)
  3. eKYC: Complete regulatory checks in AML/KYC   
  4. Onboarding: Help users learn the platform, fund their account and trade more easily and sooner 
Technical initiative add-on: Migrate the legacy registration module to a modern tech stack while keeping the current experience fully functional.

Where did AI come into the picture? 

I used common AI models to validate my discovery work and challenge my thinking.
  • First experiment: I used an AI chatbot (logged in) to simulate a discovery session. It asked questions, offered ideas, and even generated wireframes.  The output aligned closely with my strategy (Happy!), but there was a chance it was influenced by my prior chat history! 
  • Second experiment: To remove any possible pre-assumptions, redefined and structured prompts were used, and incognito mode with the same AI Agent was used, with no chat history. The outcome was less detailed, but still aligned with the core vision! Repeated this experiment by refining the prompts. Results: better detail but in the same structure as before. Not bad!
  • Third and final experiment: To validate again, used a different AI model with the same prompts as the previous attempt and again in incognito mode. The outcome echoed the same strategy! Happy!
I repeated each experiment a few times to help with accuracy, refining the prompts and Q&As. 

My takeaway from this experience: 

AI didn’t create the product vision and roadmap used in this project; it helped validate it quickly with no cost. It challenged assumptions, exposed blind spots, and gave me confidence that I was on the right track. In product discovery, this kind of low-cost, fast feedback is gold!

Any takeaway for my fellow Product Managers? 

AI can enable PMs to do a better job from idea generation to discovery, storyboarding and prototyping, from designing wireframes to developing no-code apps and do all sorts of value assessments in between.  

But AI won’t do your job (yet)! Don’t write off your product instincts, domain knowledge, or data-driven skills. Use AI as a sharp assistant or a virtual product ops partner to test, challenge, and refine your thinking. It’s not a replacement for strategy, but it’s a great pressure test.
What’s your go-to method or tool to validate product strategies? Let’s swap tips!

P.S.

  • An AI Chatbot helped polish this post, and another one was used to generate the post image!
  • I used other specialised/fancy AI-enabled tools and got the same outcome. The key to any best case of AI is having good prompts; those tools or wrappers will be handy to do this for you ( I didn't need them for my experiments) 
  • The rest of the project? That's a story for another post!

Comments

Post a Comment

Popular posts from this blog

Sell Me This Pen!

By Ali Vahed
Image
The Scene:   INT. OFFICE – DAY A sleek high-rise office with floor-to-ceiling windows. A formal interview is underway for a role in a Software development team. A table sits between the INTERVIEWER and the CANDIDATE CASTING The INTERVIEWER: Professional, relaxed. Someone who interviews for any role, from a Sales rep to Software Engineer to C-level Executive, inspired by The Wolf of Wall Street , loves asking questions that throw candidates off guard. The CANDIDATE:   Suited up, a bit nervous but composed. A seasoned software professional: Product or Project Manager, Business Analyst, or experienced Software Engineer, ready for whatever the interview throws their way.    ACTION INTERVIEWER:  (smiling confidently) Please sell me this pen! CANDIDATE  (pauses, raises an eyebrow, thinking: “Seriously? For this role?” but decides to roll with it) Sure... that’s a good question. (leans forward slightly, zoomed in on the candidate's face, thinking out loud) B...
Read more

How I approached building a 0-1 product.

By Ali Vahed
Image
What would you do if your entire product market vanished overnight? This article presents a real-world example of navigating uncertainty, validating ideas, and finding traction in unexpected places when our market shifted unexpectedly.   This experience shaped my mindset in developing software solutions from scratch. If you work in product, tech, or delivery, you may relate. THE CROSSROADS: WHEN THE MARKET SHIFTS Problem Statement: The market for our current product has been shrinking. What should we do? At Radman ITD, we created our main products based on the needs of the public sector and government departments. This drove high demand for our products, which helped us grow from a few people in a rental room to a full team with structure and our own office with a view. Then, unexpectedly, a new government administration took over, and everything changed. Our team no longer felt motivated to continue working with the new administration.  For a young company, this was a se...
Read more

Agile Frameworks for Product Managers

By Ali Vahed
As highlighted ( here ), I recently ran a workshop for our team on Agile Frameworks tailored for Product Managers. In an attempt to take my mind off everything happening in the world lately, I decided to turn the key takeaways from that session into a set of simple, visual study cards. These slides offer a quick tour through a few popular frameworks, such as  XP, Kanban, Scrum, Lean, LeSS, SAFe, etc.,  with just enough context to learn (or refresh) your understanding. Hope you find them useful! Click here to access and download the file.
Read more