Flowy One-Pager

One-line, solution-agnostic description/title.

Flowy helps non-native English speakers articulate better.

Describe the opportunity in a solution-agnostic way, such that more options could be brainstormed.

The opportunity is that non-native English speakers want to be better at communicating in English in their everyday interactions as well as professional contexts (i.e., job interviews, presentations). 

Describe, briefly, the potential benefits in the customer’s words. 

Flowy can help non-native English speakers improve their English language (e.g., clarity, vocabulary), delivery (e.g., pronunciation), and style (e.g., formality), in order to enable them to speak with more confidence and achieve their professional and non-professional goals. 

Summarize a mock customer case study involving this work (with quantifiable impacts). 

Lina is a first-year Stanford non-native student, who recently arrived in the US, is educated in a foreign language, and with low professional exposition to English. Lina needs to deliver a class speech; she feels unprepared and unconfident. She uses Flowy by recording her speech for analysis. Then, her speech is converted into text, using Flowy’s speech-to-text functionality, and she receives suggestions on how to improve her language, delivery, and style. She takes note of the recommendations and incorporates them into her speech. The outcome is an increase in the quality of her communication, an increase in English proficiency, and a boost in confidence. 

Top three areas of uncertainty, ranked by value of reducing uncertainty.

1. An AI-assisted online speech tutor that provides speech recommendations is doable.
2. Non-native English speakers would be receptive and satisfied with a voice-to-text analysis of their speech.
3. Individuals will value speech feedback from an AI.

Current plan to explore those areas of uncertainty.

Prototype Test 1 –  Implementation Prototype. We are currently building the back end of the Flowy application. Since we cannot build the entire ML model from scratch within the timeframe of the class, we plan to build the ML model for paraphrasing and vocabulary, without the connection between the back-end and front-end. However, we plan to connect the two in the future. 

Prototype Test 2 – Role prototype. We performed a Wizard of Oz prototype of the Flowy AI to find out if non-native English speakers were satisfied with the voice-to-text analysis of their expressions, delivery, and errors. 

We observed that non-native English speakers liked the personalized feedback on expressions, delivery, and errors; however, they did not like the design of the feedback response because there was no explanation for the errors. Users also wanted to get more detailed feedback on pronunciation. 

Prototype Test 3 – Implementation prototype (link). We created a speech-to-text chatbot that provides randomized scores and feedback based on a user’s speech. There is no back-end or language analysis with this implementation prototype. We also have tested different feedback formats to maximize user confidence and engagement.

Summarize prior research related to this opportunity. What do we know?

• When people have the chance to work flexibly, 87 percent of them take it. (McKinsey report)
  • This represents a tectonic shift in where, when, and how Americans want to work and are working.
• The Biggest Opportunity in Generative AI is language. (Forbes)
  • Summary: AI-powered text generation will create many orders of magnitude more value than will AI-powered image generation in the years ahead. We are seeing AI-powered text generation being applied to copywriting, contract drafting, note taking, and code development. 
• CAGR for Edtech (2022-2029, expected) 18%
• CAGR for Generative AI (2022-2027, expected) 19%
• CAGR for Language Learning market (2022-2030) 14% 

Leading signals we might observe if this is working (and not working).

• Accuracies of ML models / How Flowy compares to human evaluators
• Activation: Number of app downloads
• Retention: The number of returning customers / Churn
• Engagement: Daily Active Users (DAU)
• Monetization: Referrals

What other options did you consider (including benefits)?

We considered building an AI-enabled solution that provided feedback on non-native English speakers’ everyday conversations. This would involve recording everyday conversations and providing after-the-fact analysis. Although this was potentially impactful, we decided to not do it because the user experience would likely involve a lot of friction and raise privacy concerns. 

We considered building an education technology platform for non-native English speakers. However, we decided not to adopt an edtech approach because based on our customer interviews, non-native English speakers want to have the option of being task-oriented (i.e., getting feedback for a presentation) and learning-oriented (i.e., learning new vocabulary). Therefore, Flowy is positioned as a tool, not an ed-tech platform. However, there is an opportunity for users to learn from the feedback that Flowy provides, enabling users to improve their English proficiency. 

Why now, compared to alternatives? Really, truly, why now? Cost of delay?

Most importantly, there has been a big shift towards remote work, and telecommunication has increased 2.6-fold.

From a technological point of view, Open AI recently released Whisper, a multi-language language model with 1.6B parameters. Furthermore, several startups are also starting to shift their attention towards this problem space. 

Brief summary of technologies involved (if known). Detail any areas of specialization required.

• Speech Processing – For speech-related features
• Natural language processing – For paraphrasing, and feedback generation
• Full stack development – For app development

Rough time frame (e.g. 1-3w, 1-3m, 1-3q).

1-3 weeks:

• Build an NLP algorithm with no back-end to address risky implementation assumptions
• Iterate on UX/UI to optimize voice UI 

1-3 months:

• Finalize Flowy’s brand identity and UX/UI 
• Implement back-end development 
• Build NLP algorithms for must-have features, such as pronunciation, vocabulary, and paraphrasing 
• Conduct user interviews and test works-like prototypes 

1-3 quarters:

• Ship out Flowy’s beta version
• Receive venture funding