[Editor's Note] OpenAI CEO Sam Altman once said at the YC alumni meeting that "companies that simply package OpenAI are destined to not survive in the long term." As OpenAI continues to bring updates to its large models, the era has come when everyone can easily create GPT, and AI startups that previously innovated based on this application dimension have also been greatly impacted. So in the era of big AI models, what kind of AI startups will have a way out? In this regard, Decibel VC partner and Databricks investor brought his thoughts.
Original text: https://dannguyenhuu.substack.com/p/defensibility-in-genai-what-ai-startups
Translation Tool | ChatGPT
Editor | Su Mi
Produced | AI Technology Basecamp (ID: rgznai100)
The following is the translation:
Earlier this month, OpenAI launched products at its first developer conference, which caused strong reactions from many startups, which also triggered poses an important question: How can startups gain a foothold in an era dominated by fast-growing incumbents?
It's reminiscent of the rapid innovation that sparked in the mid-2010s when AWS dominated cloud computing. Their influential re:Invent keynotes often cause startup founders to reconsider their reasons for entering the cloud computing space. However, we still have large infrastructure companies like Datadog, Elastic, and Databricks that were all founded during that period.
It seems that the situation will be similar this time.
Shortly after OpenAI’s Dev Day and just a few blocks away, GitHub unveiled new features for Copilot at its GitHub Universe conference. With over 1 million paid users from over 37,000 companies, GitHub Copilot as a software-as-a-service product provides an interesting case study in how to build and sustain long-term differentiation while using OpenAI’s foundational LLM.
While Copilot itself is part of the larger company Microsoft, many of their strategies can be applied to startups and scaling companies trying to build AI products. In what follows, I try to highlight some strategies that I think are very transferable, and also aim to provide interesting guidance for founders currently building companies in the AI space.
Community = first-party data, dissemination and adoption
Participating in the community is becoming increasingly critical to branding an AI-driven company. Vast amounts of data generated from open source or freemium users, combined with rapid feature requests and feedback, are critical to building an efficient and reliable AI service.
In recent years, open source companies have primarily monetized their user base through cloud-based managed services or enterprise features such as role-based access control (RBAC), high availability, and enhanced security. This approach typically involves charging in order to minimize the infrastructure effort required to effectively deploy the service within the organization. For example, Elastic operates Elastic Cloud, a managed service version of its open source product, while MongoDB takes a similar approach with MongoDB Cloud, among others. This “convenience layer” monetization strategy essentially frees initial community users from concerns that are irrelevant to them but are necessary for enterprise requirements.
Looking ahead, companies with strong community bases are likely to leverage their vast data to create new services using LLMs and agent technology. These services are complex both technically and from a use case perspective. GitHub's Copilot is a typical example:It uses repository data to train its initial version and continuously improves it, putting its core users-developers at the forefront. Every interaction in which a user accepts or rejects a code suggestion provides them with the opportunity to collect first-party data, which not only continuously improves the product, but also creates a long-term defensible mechanism. Considering Copilot's complexity and utility, creating a tool that developers not only use but love is an important achievement.
The strategic advantages of having a built-in community are clear, providing users with opportunities for data collection, implementation of best practices, dissemination, and product testing. For those modern service-as-a-service companies that are lucky enough to have such a community, the challenge and opportunity is how to balance complexity and convenience within a single coherent product strategy based on first principles.
In short, if GitHub were just starting today, how would it jump directly to Copilot?
Fascination with UI/UX + AI = User Happiness
Previously highlighted on GitHub Universe The latest developments in GitHub Copilot provide insights for AI startups looking to build lasting and successful businesses Important insights. Their deep integration of UX and UI in AI products, such as GitHub Copilot Chat and Copilot Enterprise, is a testament to balancing sophisticated AI capabilities with user-centered design. For AI startups, this means not only leveraging advanced LLMs but also embedding it in an intuitive way that matches user expectations and workflows.
GitHub Copilot Chat, powered by GPT-4, demonstrates their commitment to UX/UI by enabling natural language programming, providing code-aware guidance, inline chat for specific lines of code, and user-friendly shortcut commands for tasks. Because it's integrated into GitHub.com and its mobile apps, this tool is not only highly accessible, but also enhances the developer experience by providing suggestions, summaries, analysis, and answers to coding queries directly within the platform.
Copilot Enterprise edition tailors the Copilot experience to organizational needs, providing teams with AI assistance at every stage of the software development lifecycle. It provides personalized code suggestions and documentation assistance to quickly familiarize teams with their specific code base. This customization, coupled with enterprise-grade security and privacy features, illustrates how GitHub Copilot has evolved from a simple autocomplete tool into a comprehensive, AI-powered development aid.
AI startups hoping to emulate the success of GitHub Copilot should prioritize creating AI tools that are technically sophisticated, intuitive, and deeply integrated into user workflows. This focus on user-centered AI development, coupled with continuous innovation, forms the cornerstone of building a long-term viable business in the highly competitive AI industry.
Deep integration into multiple core systems
AI startups can gain important insights from GitHub Copilot’s recent preview release of “Workspaces.” This move fully reflects the strategy of deep integration into various systems to create a unified and flexible user experience.
As I previously posted, "Many-to-Many Problem (Many-to-Many Problem, https://dannguyenhuu.substack.com/p/a-new-frontier-service-as-software) As described in the article, the ability of artificial intelligence agents and LLMs to navigate the system maze, integrate and interact between tools, can greatly improve operational efficiency and the effectiveness of decision-making processes.
GitHub Copilot's move into Workspaces highlights this potential, addressing the complexities of modern software development environments. By leveraging knowledge of the entire codebase and the inference capabilities of GPT-4, GitHub Copilot Workspace helps developers efficiently turn ideas into code. This not only streamlines the development process but also integrates all aspects of software development into a cohesive workflow.
Creating a long-lasting agency system depends on integrating systems, workflows and data. GitHub Copilot Workspace embodies this approach, demonstrating how AI can be leveraged to solve specific use cases while aligning with existing enterprise budgets and needs.
For AI startups looking to build sustainable business models, GitHub Copilot's strategy for scaling and integrating across multiple systems provides a valuable blueprint. By focusing on deep integration and solving specific user needs, startups can create products that not only solve immediate problems but integrate seamlessly into broader workflows, thereby building long-term sustainable competitive advantage in the software-as-a-service space.
in conclusion
Building a durable and defensible position in the rapidly evolving world of GenAI requires startups to leverage their unique strengths. This may include:
Cultivate and leverage a strong, engaged user community
Emphasis on combining user interface and experience with artificial intelligence to meet user needs
Develop deeply integrated AI agent systems that span across core systems to provide key insights and operations
Ideally, a combination of these methods will be most effective. While OpenAI's rapid innovation within its ecosystem poses challenges for startups, the strategy adopted by GitHub Copilot to become a leader in AI-powered software development tools provides valuable lessons.
There's no doubt that Copilot clearly benefits from being part of Microsoft, yet these strategies still provide an interesting blueprint for AI startups to establish themselves as a strong presence in this new era.
