Understanding Competition in AI Chatbots

OpenAI launched #ChatGPTEnterprise yesterday. This is a game-changer in many ways. It puts several startups at existential risk and challenges its larger competitors. Let's dig deep. Here are the added Enterprise features that come out of the box with ChatGPT Enterprise. 1. Enterprise-grade security and privacy (SOC2, SSO, Central Admin Console, Usage Insights) 2. Unlimited higher-speed GPT-4 access 3. Longer context windows (32k window) 4. Advanced data analysis (formerly known as code interpreter) 5. Collaborate and build custom workflows (shared chat templates) So, what does this mean to OpenAI's competitors and the several thousands of startups building on LLMs? Who all would get hit? LLM Providers 1. #Claude was positioning itself as the 'Enterprise Ready' LLM provider, but OpenAI's early adoption (80% of Fortune500 uses ChatGPT in its current form) by enterprises is super hard to beat. 2. And having worked on Meta LlaMA2 models, I can say that #LlaMA2 is still far-off when it comes to OpenAI's GPT models. #HuggingChat has several LlaMA models available to test and you can tell the difference quite clearly. GPT Wrapper Companies 1. Hundreds, if not thousands, of startups (including several YC23 batch startups) are nothing but a wrapper on GPT. Even the outgoing version of #Lyzr, the product we are building, was nothing but a better UI/UX retrofitted on GPT. 2. LLM wrappers in Analytics, Data Science, LLMOps, LLM admin, Chatbots, Document Search, Prompt Libraries, Data Lake Search, Summarization, Copy Writing, Basic Research, Email Writer, Custom Workflows, ChatGPT Collaborators will be asked this question again and again - "So, how are you different from ChatGPT Enterprise?" Yes, in the defense of the 'wrapper' startups, we can integrate more LLMs like #Claude #LlaMA2 #Bard into the app and give users the choice like #Huggingface does. But honestly, how many of us really use the several thousands of assets available in Huggingface? There are massive discussion threads on Reddit & Twitter discussing the unusability of Huggingface assets as it is just overwhelming and not super helpful. So, what should startups do? Well, I am still trying to figure out the answer. But it would be foolish to bet against OpenAI's Enterprise focus (powered somewhat by Microsoft, too) and continue down this path, assuming that OpenAI will eventually stop building stuff in this space. OpenAI is to LLM startups what AWS was to enterprise startups. 

Semrush recently published a stat that really caught my attention: 5.5% of ChatGPT’s outbound traffic goes to Google, while only 4.5% of Google’s traffic is heading back the other way and that number is down 33% month-over-month. At first glance, this might seem like a small shift. But it tells a much bigger story about how search is evolving and where users are actually staying. What we’re seeing now is a subtle power shift. ChatGPT still feeds on Google’s data. But once users enter that interface, they’re not bouncing back to Google. They’re staying inside the AI experience, getting what they need without ever opening a new tab. Meanwhile, Google created AI Mode in an effort to adapt. It’s not replacing classic search entirely (yet), but it’s clearly signaling a new hybrid model: part AI, part traditional. Discovery. journey. is. fragmenting. Users are asking. Exploring within AI tools.  And increasingly, they’re not returning to the search engine, they’re getting everything they need right there in ChatGPT, Perplexity, or Gemini. So even if your website ranks #1 on Google, you may be invisible in an AI-generated answer. The answer? Thinking about representation in the AI layer. And understanding where user intent lives now and meeting it there.

Character.ai, an AI app enabling users to design AI personas, has seen a surge in U.S. mobile app usage, reaching 4.2 million monthly active users since its May 2023 launch. In comparison, ChatGPT reports nearly 6 million monthly U.S. actives. Notably, Character.ai has captured a significant portion of the younger demographic, with 60% of its web audience coming from the 18-24 age bracket. ChatGPT’s traffic from this group declined to 27% in July. These apps are quite different from one another in their approach to conversational AI, but there may be something to learn here about how to package and market AI apps. On the global front, ChatGPT maintains a lead with 22.5 million monthly active users, overshadowing Character.ai’s 5.27 million. While ChatGPT remains dominant on the web, Character.ai’s users show a marked preference for mobile interactions. Character.ai’s recent growth is backed by a substantial $150 million in Series A funding, valuing the startup at $1 billion. The company was founded by Noam Shazeer and Daniel De Freitas, known for their contributions to Google’s LaMDA project. 'Character.ai Challenges ChatGPT’s Dominance in U.S. AI Market' #AI #ArtificialIntelligence #chatgpt4 #GenZ