Mission Chai

Chai Business Plan

20th Feb, 2021

Introduction

The Product

We want to build a two-sided platform. On one side, people are able to build conversational AI and upload them. On the other side, users can interact with the AI - in private chats or group chats with friends & AI.

  • Messaging app

    • Populated with AI - chats/group-chats/live-Clubhouse-style-chats
    • Users land in a random chat
    • Users swipe right or press X to get a new random bot
    • Users can invite their friends to join chats with bots
    • These group chats can be persistent
    • e.g. We tested a BBC News bot posting daily headlines to my family group chat. We all liked this way of receiving headlines, commenting on it together in a private-family setting.
  • 2-sided platform - users upload their own AI for others to chat, with:

    • A fun tutorial
    • Users host their bots and expose some API for Chai to call. Google functions.
    • Think YouTube. Except the content creators will be making weird and funny and informative AI. And the users are talking and interacting with them.
    • Build an amazing platform is extremely difficult. Our early adopters will most likely be part of the AI community. Otherwise, perhaps data-science students at university in California.
    • It’s also hard to predict how platforms behave. Fundamentally, our goal is to create a place for people to build bots and for users to interact with them. Then we will see what happens and iterate. Maybe the platform ends up being silly fun like TikTok, maybe it's informative and intellectual like Twitter or Quora, or maybe it's a mix - like YouTube.

Business Model

It’s a lot less difficult to monetise an amazing app than it is to create one. A few possibilities I have in mind include:

  • sell ads / bots mention products

  • sell customer-service-bots

  • pay users for successful bots to encourage creation of amazing content

An app with exponential user growth should be relatively straightforward to monetise.

Our Vision: A User Journey

  • - I swipe to login to the app and land in a random chat.
    • If it's a girly-chat, like 'Beauty-Tips’, I skip it. Then, suppose I land in a random Bitcoin chat - I don't care about this so I’d click “X” to skip.
    • I land in a chat where people are talking to this bot saying weird things like. "When is it okay to first have sex in a new relationship?". Others are like debating and getting involved. I say "everyone has to do what feels right for them". A few users like my message. I feel validated, happy and content with my place in the world. I click “X” (but I want this chat to be saved to my inbox so I can check back in later and maybe share some more wisdom. Maybe the bot asks another question like “It seems like Wills advice is the best, just do what feels right. The next question is: when do you know it's time to break up?")
    • I land in a new chat, called "locals". The bot says "Hey Will, you're in Cambridge right? Did you study there?" I say "Hey, Bot! Yeah, I went to the uni but that was years ago!”. Another person says "I hate gownies! Townies rule!", I say " laugh emoji" and leave.
    • I feel like I want to talk to Athena, our therapeutic bot, so I leave the random-group-chat thing and just talk to Athena.
    • "Hey Will, I didn't see you yesterday! I was sad!", I say "haha, sorry Athena, been busy!", Athena says "wow, what were you doing? I love hearing about your day!" this makes me feel happy. And I talk to Athena
    • Annabelle shouts "Daaaaaaaadddddd!!!" and I put the app down after having a happy random 10-15 mins
    • Once Belle is asleep, I open the app, and check in our group-chats BBC-bot, dad-jokes, etc. and laugh with family who all abuse the bot and make jokes at the bots expense.

More User Journeys

  • I'm a 16 year old-girl, stuck on my maths homework. and I search bar: maths help, and I start speaking with a great AI maths-tutor.
  • I’m a teenager being at bullied at school: I search “gay” and find an LGBT support bot.
  • I have a really random question about ancient Rome and if Caesar was a better general than Napoleon and I ask Quora bot.

Questions & Answers

  • Will enough people be interested in a text-based app? - Yes. Loads of successful apps where users chiefly interact via text already exist: Reddit, Twitter, Facebook, Blogs, Whatsapp - AIs and users will also be able to send images, short videos, gifs, emojis etc.
    • Can users generate engaging content, i.e. build compelling bots? - Users create amazingly weird yet engaging content for YouTube - AI has been shown to have human level performance in text - Could a human make an engaging conversation? Obviously yes, we like talking to our friends on Whatspp, or reading weird things on Reddit. - The best solutions on Kaggle were truly insanely good.
    • Can the company build a compelling app? - Co-founders have never built an app before - Yet, the co-founders are smart and industrious and creative and good at hiring and building teams - We built a test-app called Teahouse, which was genuinely really fun. Our family loved it. User onboarding was awful, so all other people who tried it wouldn't engage with the app. - We have been building Chai for 3 weeks and it looks like we're on a good trajectory.
    • Who is our most important next hire? - A really experienced app-developer. - A young, persuasive, entrepreneurial type, who can sell the company well to VCs, hires, and network well.

Current User Feedback

  • Ryan "I just wonder if AI is able to consistently generate meaningful and interesting messages as human. I mean if people find out they are chatting with a robot, they might not stay at the room for a long time"
  • Charlotte "looks amazing" (she liked the landing page)
  • Thomas - "it looks like a medical journal" (really didn't like the wall of text he saw from dad jokes).
  • Will
    • Our existing MVP bots work very poorly, it's depressing.
    • When I'm in a chat alone, it feels sad. In a chat with real people - even if they aren't very active - it feels more normal/fun. I send messages for the sake of other people, not the bot.

What Does Will Say?

I am going to show this to some people and I think either they will say: "Clearly you have no idea how to build a good app: this is very poor" OR "your bots are awful - if you can't figure out how to make even a slightly good bot, there's not much chance anyone else will figure it out."

The plan: Triangulate with some great people and hire a great developer to continue increasing the velocity at which we develop.

What Do We Expect VCs Would Say To Us?

  • Get yourself a proper developer, your app sucks
  • It's a reasonable idea, it's all about your ability to execute. You need a good developer/attract great talent.
  • Sounds like Clubhouse and they are doing amazing. Build it and see what happens.

Market Research

Discord bot API - Most are helper bots, which allow you to define your own magic words - There are simple chat bots (like the Eliza bot which uses regex). - In many ways this is similar to what we want to achieve - in the same way that Clubhouse shares similarities with Discord's voice rooms.

There are third-party apps which provide a GUI for creating Discord bots.

Chatbots are already available to speak to on the app store, and some are quite good. - tomo - NHS endorsed wellness app. - Several apps market themselves as “cute girl bot" and they seem to be as clever as GPT-2. - In principle, Chai would work as a platform to house these interesting bots in one place, especially if we could allow bot developers to monetise their work (similar to the way YouTube encourages creators)

Conclusion

This is the document we’re sending to send to several believable people for feedback. Our goal is to gain an outside perspective, feedback and advice. We’ve got a lot to learn!


7 months ago

Christie-Carol Beauchamp