Google's Strategic Move: Betting Big on Character-Driven Chat Experiences

Recently Reuters reported that Google was in talks to invest "hundreds of millions" in Character.AI at a $5B+ valuation. This is actually a pretty big deal, for both Character.AI and for Google.Character.AI is an AI platform known for creating interactive, conversational AI characters. These characters are designed to engage users in realistic and dynamic dialogues, simulating human-like conversations. Basically it allows users to have conversations with different personas or avatars as if the user is really talking to that persona.For example, a user can have a conversation with Alexander the great, George Washington, or even create their own character such as a pirate, or your company's persona and users can chat with them as if it was a real conversation.

Conversation with Alexander the Great on Character.AI
Conversation with Alexander the Great using Character.AI

What value does this bring?On the surface, you may think that is nice, but it's a toy. What real value does that bring? You are right to a certain extent. The primary use of Character.AI is entertainment. But, the technology they are building can be repurposed for other uses, which is why I believe Google made the investment.Here are a few use cases where this would work.


Use Cases

Chatbots/Customer Service - The first use case is to take your company's brand persona or to provide each user with their own persona for your company. I've been in the conversational AI space for some time. When working with many of these companies, the thing they want most after accurate answers and automating the transaction, is to do it in their voice.

If you talk to the marketing leader at any company, they will tell you how hard they've worked to build their brand voice. Some companies choose to be casual and friendly while others may choose to be more buttoned up. Building this into their chatbot has traditionally proved to be difficult as it required so much time to generate content specifically for their bots and the ability to actually test it according to this requirement.

The other side of the spectrum is to have a persona catered to the user. One of the values of generative AI is having a segment of 1. Having the same brand voice for all customers, regardless of their demographic is not always the best thing. If I'm a 25 year old working with a wealth manager, I may prefer interacting with someone that understands me, my likes and interest, and speak to me in words that I can understand. On the other side, if I'm adept at investing, I may want someone who uses more complicated terms and is more direct. While Character.AI may not have these features today, They have some of these capabilities built out internally, because they would not have the ability to provide the service they do with out being able to test, switch out, and evaluate these without it. It's just a matter of exposing it.

Marketing - building personas and talking to your customers before actually talking to them. Many customers build out user personas when building new products or services, or when they are going to launch a new campaign. Companies can build many of these personas and have conversations with them. Something as simple as asking the persona to play devil's advocate can help identify things you didn't think about. Roundtable is doing something similar by allowing you to leverage your data and AI personas to simulate surveys.

Emotional support - while the primary use case is chat entertainment, people are using Character.AI for emotional support. They are having conversations with their bot that essentially becomes their friend. The technology can offer this emotional support through AI characters that are available 24/7, providing non-judgmental interaction and stress relief, which can be particularly beneficial for discussing sensitive issues or practicing social skills. This same use case can be exposed via partnerships with companies such as Calm for sleep health or Headspace for mental health.

Education plus Learning & Development - How cool would it be to talk to a historical figure and learn directly from them. Character.AI can be used to build out these interactive and experiential education. On the other end of the spectrum, it can be used for professional training such as watching how your employees deal with belligerent customers or teaching your hospital staff how to deal with different patient scenarios in medical training.

Agents and Prompt Building - when building agents or even simple prompt based applications, you typically start with a persona. These typically start with, "You are a {persona type}", where persona type is the role such as "you are a doctor" or "you are a marketer". These types of openings prime the model to have a strong probability of predicting words that are within that space. Here are code examples where you see a list of roles and some of the actions followed by an abridged example prompt for a software architect persona.

Abridged prompt example using a Software Architect Persona

More entertainment - This one maybe a long shot, but what if they can extend this to build out characters for shows, movies, and games. Entertainment companies can build catalogs of characters and then allow writers to pick and choose or remix characters as part of their creative process.


There is so much that can be done in this space. Much of the examples described assumes Character.AI will pursue other paths beyond consumer into the enterprise space or at least partner with other companies that will, such as Google. To support this, Character.AI will need to build out several capabilities that ease the engineering, testing and evaluating, and managing these personas. This is basically why I believe Google invested. Google has some commercial interest across each of these use cases in some form. While on the surface it seems like fun chatbot experience with historical figures, we're really peeking into one of the possible futures with AI. It's still early days, but if Character.AI plays its cards right, we could be watching the rise of another AI platform. Bottom line, I would say watch this one closely.