Do you actually end up talking to users first?
Soumil Rathi
19 replies
I've always found it tougher to talk to people before building something, just because you don't have a good idea of what you're building and for whom.
Personally, I tried solving this by building a project to simulate those initial customer interviews with AI. I'm launching this soon, if you're interested, but was curious—what do you do now to solve this problem?
Replies
Nick from FirstHR@nickanisimov
FirstHR
Any AI will help you prepare for a conversation now, based on your tasks and situation, AI will easily make you a conversation plan and a list of questions.
But the most important thing is to start. It's often scary and may not go very well at first, but then with each call it will get easier and better.
Share
I try to have casual conversation with potential users early on. It's less intimidating and often gives surprising insights.
Hi, I'm starting a new project soon and I will start by talking to users first but they are all people I know.
I feel like that is going to make things a bit easier for me.
Do you have any advice on how to do this?
Launching soon!
@luka_brzin I think that can work, but you need to ask non-biased and non-leading questions, especially considering that these people are likely to be positive towards you. Try to not mention the solution and try to figure out what problem the customer actually wants to solve.
I'd recommend reading the Mom test and reading about the Jobs-to-be-done framework. Both of those have helped me a lot.
I've struggled with this too. Sometimes, reaching out to people who already face the problem helps clarify the next steps.
Recap
Try to relax a bit and approach conversations without a clear agenda. Reflect on them afterward. Sometimes, you might be pleasantly surprised.
Perhaps an unpopular opinion, but indirect insights are also just as helpful in the super early stages while you're still forming your idea. It's 2025, people aren't shy of writing (Reddit) and talking (TikTok) about their experiences, both positive and negative. A quick keyword search around the general topic you're pursuing will pull up thousands of people expressing their unfiltered thoughts and pain points.
Then, once you see a general pattern or trend, refine your idea and direction and share your idea with a few users. Iterate and repeat as needed.
I think the problem you've identified is very real and the solution seems super neat! Curious what you're using to train the AI to mimic the precision of actually talking to users who's experiencing the problem - would be really cool if it somehow was pulling from social media as mentioned above.
@roy_khan2 there's a huge underrated benefit to it too. It removes all underlying bias variables that you'd have to be cautious of in a live user interview (e.g., there's no leading questions that you may unintentionally ask that could skew answers).
@carol_leung I love your suggestion about mining social media platforms like Reddit and TikTok for unfiltered insights; it's a great way to spot patterns and trends before engaging directly with users.
Launching soon!
@carol_leung Hi! For now, I'm using multiple AI agents with different personalities and information to simulate a conversation.
I've personally used social media a lot for refining my idea. So, I agree that the ideal solution would integrate the AI with any insights and preferences picked up from social media—that's probably something I'll look into doing when expanding the project!
Starting with a vague concept and refining it during conservations has worked for me. Users often shape the idea better than I can alone.
Launching soon!
@morgan_harriss Yeah completely agreed. The best way to refine and come up with an idea for me has been to talk to people about it and look at feedback.
I usually start with small surveys or polls. It's less direct but still gives me a sense of what people want without overthinking
Launching soon!
This is such a relatable challenge, Soumil. Talking to users early on can feel intimidating, but it’s so critical to building something meaningful. I like the idea of using AI to simulate those initial customer interviews—it seems like a great way to ease into the process and refine your approach before diving into real conversations.
Launching soon!
@roy_khan2 Thanks! It's mostly intended for those who don't know who to reach out to, and those that aren't ready for those real conversations yet
Soumil, how does your AI simulate user behavior and feedback? Is it trained on specific user personas or more general patterns? I’d be curious to know how it balances precision with flexibility for different industries
Launching soon!
@roy_khan2 Hi! For the most part, I've used a multi-agent system (customer, interviewer, evaluator) to simulate certain personas and real life experiences. Because LLMs have so much information about any industry, it's able to be flexible but also give pretty actionable advice.