StepOne

Inspiration

Over 90% of startups fail, with the lack of market need cited as the #1 reason for failure (CB Insights, 2018). At the heart of this issue is a broken or non-existent customer discovery process. Despite the popularity of frameworks like Lean Startup, most founders still struggle with the how of effective customer interviews.

Today, with AI the technical barriers to developing applications and integrating systems have been greatly eliminated. However, we believe the greater challenge now lies in understanding market demands and aligning with what customers truly want.

We recognized that founders often enter interviews with biases, ask leading questions, and hear only what they want to hear. As a result, they build products based on assumptions—not reality. According to Batova et al. (2016), this overconfidence and lack of qualitative rigor in lean startup customer discovery creates a "validation void" that can doom ventures early. So we decided to build something to aid customer discovery for the next generation of builders.

What it does

Our solution is an AI-powered Customer Discovery Agent. We offer a smart, interactive platform that simulates real customer conversations. Founders can pitch their ideas to diverse AI personas (like Sarah, a single mom), engage in guided interviews, and receive real-time analysis of pain points, competitive threats, and unmet needs.

Our Three-Step Approach

Pitch Your Idea: Founders describe their startup idea and select relevant personas.

Practice Interviews: Natural language interviews help founders learn to ask open-ended, unbiased questions.

Insights & Feedback: A reasoning agent provides an instant summary of key insights, highlights areas of concern, and offers suggestions for better questioning.

This is more than just practice - it's a training ground that replicates real discovery conditions, while helping founders break out of their confirmation bias loops.

How we built it

Our project architecture includes the following key agents:

Interview Agent: Realistic AI personas powered by language models.

Coach Agent: Offers tips, critiques, and suggests follow-up questions.

Market Research Agent: Pulls contextual data (news, product recalls, sentiment analysis).

Orchestrator: Coordinates all agents and manages the feedback loop.

The platform leverages LLM APIs for persona dialogue, custom prompt engineering for interview realism, and backend data pipelines to extract and synthesize key insights from each conversation.

Challenges we ran into

Avoiding Leading Questions: Just as real founders struggle to avoid leading the interview, we had to train our Coach Agent to detect and flag such behavior. This required curating hundreds of bad vs. good question examples.

Accomplishments that we're proud of

We're proud of is how our system empowers founders of all backgrounds to conduct customer discovery with confidence. Integrating these different agentic systems, and ensuring they behave as reliably as possible was a challenge....

What we learned

Customer discovery is a skill, and like any skill it can be taught and practiced.

Lean Startup methods often lack practical guidance on how to conduct interviews (Batova et al., 2016).

Many startups focus too much on building and not enough on validating. By integrating learning loops into discovery practice, we can help founders build what people actually want.

This experience also reinforced our belief in the potential of AI not just to automate tasks, but to coach and educate entrepreneurs in ways that scale.

Our Vision - What's next in the pipeline :)

Real-time suggestions for what questions to ask next (leveraging feedback agents that monitor conversations and provide timely prompts).
Expanded persona library featuring a wider diversity of demographics, accents, and conversational styles to ensure more inclusive and relatable interactions.
We wish to add support for video interviews, allowing users to practice in a face-to-face setting.

References

Customer Discovery Basics and Best Practices Darren Cooke: https://www.youtube.com/watch?v=sW61nRfNjgg

Batova, T., Clark, D., & Card, D. (2016). Challenges of lean customer discovery as invention. 2016 IEEE International Professional Communication Conference (IPCC), 1-8.

York, J. M. (2020). What is lean startup: A solution to startup failures or not? International Journal of Tourism & Hotel Business Management, 2(2), 250-264.