Key Moments
How To Use AI In Your Startup
Y CombinatorKey Moments
AI isn't a magic bullet for startups, but building with LLMs at the core is essential, and embedding in the Bay Area's AI community accelerates innovation.
Key Insights
Founders should consider what will be possible when current LLMs are twice as good today, rather than just focusing on current capabilities.
Leveraging AI internally for efficiency is no longer optional; it's as fundamental as using the cloud was in the early 2010s.
Companies like Superpowered and Bappy demonstrate successful pivots and growth by deeply embedding in the AI community and adapting to technological shifts.
Pivoting to AI without novel insights or changing one's environment and customer engagement often fails, even with simple OpenAI calls.
The Bay Area offers unparalleled proximity to AI expertise, enabling rapid learning and iteration that is difficult to replicate remotely.
The US healthcare system's $1.3-$1.4 trillion in administrative spend, largely due to legacy systems and human data transfer, presents a massive opportunity for AI automation.
Why building with LLMs at the core is essential
The exponential pace of Large Language Model (LLM) improvement necessitates that startups not only consider but integrate AI at the heart of their operations. This is not merely about adopting a new technology but about fundamentally rebuilding existing software or processes with AI natively embedded. Just as the early 2010s saw a wave of cloud-native software replacing on-premise solutions, today presents a similar opportunity to create superior products by leveraging AI from inception. While a direct pivot to 'AI' as a buzzword is ill-advised without genuine innovation, any new company today should consider it nonsensical not to leverage AI. This includes using LLMs for internal efficiencies, such as in HOA management companies that streamline operations, or for customer-facing products that offer entirely new capabilities. The question for founders should be: what will be possible when these models improve 2x, and how can we build for that future now?
The cyclical nature of technological waves
Experiencing major technological shifts, such as the mobile wave following the iPhone's introduction in 2007, offers valuable perspective for current AI trends. Founders who lived through these cycles can recognize similar patterns and opportunities today. In the mobile era, it took time for app ecosystems and permissions to mature, but the subsequent five years saw the explosion of mobile companies. Similarly, the AI revolution, which has been intensifying over the past two years, is creating a fertile ground for startups. Big companies, while eventually adopting new technologies, are often slower to execute than nimble startups. ChatGPT, for instance, has been around for over two years, yet there's still significant room for innovation and improvement – the analogy of a 'dumb Alexa' persisting despite obvious direction highlights this.
Pivoting successfully in response to AI
While simply calling OpenAI doesn't guarantee success, strategic pivots driven by a deep understanding of AI's potential are crucial. The story of Superpowered and its evolution into Bappy serves as a powerful example. Initially a financial investment platform, Superpowered pivoted during COVID to a productivity tool for Zoom calls. This demonstrated an ability to adapt to market needs. Later, recognizing the AI wave, the founders made a decisive move to San Francisco and shifted focus again, leading to Bappy's rapid growth as a key player powering many YC companies. This success wasn't accidental; it involved embedding themselves in the burgeoning AI community and proactively shifting away from a previous product before it was obsolete. This proactive adaptation, rather than a reactive 'pivot to AI,' is key.
Avoiding common pitfalls in AI pivots
Many startups falter when attempting to pivot to AI because they lack genuine insight or a compelling new approach. Simply positioning oneself as 'customer support agent company number 50' and making calls to OpenAI won't change a startup's fate. Success requires more than superficial integration. Founders must gain deeper customer insights, rethink workflows, and potentially change their operating environment. Without these fundamental elements, a superficial pivot is unlikely to yield results. The core of building a successful startup – delivering value to customers – remains paramount, regardless of the technology employed. AI should enhance, not replace, these foundational principles.
The strategic advantage of the Bay Area AI ecosystem
Relocating to or temporarily immersing oneself in the Bay Area offers a significant competitive edge for AI startups. This region has become a dense hub of expertise, innovation, and community. The ability to walk down the street and connect with world-class talent or companies working on related problems, as exemplified by the comparison to learning SEO from Pinterest when they were geographically accessible, is invaluable. This proximity facilitates rapid learning, idea exchange, and a clear understanding of the state-of-the-art that is incredibly difficult to achieve remotely. Experiencing this environment firsthand provides crucial insights into what's feasible, what's impressive enough to close deals, and what gaps still exist, preventing founders from being unknowingly behind.
Automating AI for specialized domains like healthcare
AI, particularly LLMs, is poised to revolutionize specialized industries by automating previously complex or labor-intensive tasks. Companies like Replex (AI localization) and Gecko Security (AI security engineering) exemplify this by transforming niche skills into accessible software. In healthcare, a sector with immense administrative overheads (estimated $1.3-$1.4 trillion in US spend), AI offers vast potential. Many healthcare administrative tasks involve manual data transfer between legacy systems, creating inefficiencies and higher costs. Companies are emerging to automate these processes, such as generating pre-authorization requests or summarizing patient information for billing. The challenge for founders targeting healthcare is often a lack of direct experience with these specific inefficiencies. The advice is clear: embed yourself in the domain, observe workflows, and identify repetitive tasks ripe for automation.
Deepening AI impact through direct customer engagement
The ultimate success of AI applications hinges on solving real customer problems, whether in B2B or B2C contexts. For areas like healthcare, beyond automating back-office tasks, AI can directly aid patients. Examples include AI-powered voice systems that check in on patients between visits, monitor their wellbeing, and facilitate necessary follow-up appointments. This not only improves the patient experience but also streamlines practice management, creating a win-win scenario. The core principle remains: founders must understand the end-user's needs deeply. Spending time observing how professionals in a given field perform their tasks – sitting next to them and watching their screens – can immediately reveal numerous repetitive tasks that AI can improve or automate, yielding significant value.
Mentioned in This Episode
●Products
●Software & Apps
●Companies
●Organizations
●Concepts
●People Referenced
AI Application Examples in Startups
Data extracted from this episode
| Company/Area | Problem Solved | AI Solution | Impact |
|---|---|---|---|
| HOA Management Company | Internal Inefficiencies | LLMs for efficiency | Improved internal processes |
| Superpowered (early), Bappy (current) | Zoom call management (early), AI-powered business solutions (current) | AI integration, pivot strategy | Rapid growth, powering many companies |
| Replex | Manual UI Translation | Automated AI localization | Streamlined translation |
| Gecko Security | Specialized Security Skills | AI Security Engineer | Automated security management |
| AI Co-pilot for Medicare Advantage Agents | Complex Insurance Workflow | AI co-pilot | Assisted sales process |
| Tara | Manual Pre-authorization Requests | LLMs for summarization and request generation | Automated administrative task |
| Voice AI for Patient Outreach | Patient follow-up and scheduling | Voice AI calling patients | Improved patient experience and practice business |
Common Questions
Yes, you should almost absolutely be working on something that uses LLMs at its heart today. However, simply making calls to AI models without fundamental value creation for customers won't guarantee success. Focus on leveraging AI to build better products or improve core processes.
Topics
Mentioned in this video
Large Language Models are discussed as a core technology for new startups to leverage, offering efficiency and new product possibilities.
An example of a business using LLMs internally to make operations more efficient.
The specific type of insurance plan for which an AI co-pilot was built, assisting agents with a complex workflow.
Discussed as a massive sector with significant administrative costs, ripe for AI-driven efficiency improvements.
A consumer productivity product built during COVID that helped manage Zoom calls, later evolving into Bappy.
Mentioned as a provider of AI models whose calls can be integrated into startups, but simply calling them isn't a guarantee of success.
A company using LLMs to automate the creation of pre-authorization requests in healthcare, streamlining a manual administrative task.
A company acting as an AI security engineer, automating specialized skills with AI and an intuitive interface.
The original name of a company that pivoted from a financial investment platform to a productivity tool (Superpowered) and later to Bappy.
A company that pivoted from Superpowered and is now rapidly growing by powering many companies within and outside of YC, leveraging AI.
A company building software for automated UI translation (AI localization), showcasing a specific application of AI.
The previous base of the Superpowered/Bappy team, from which they decided to move to San Francisco to better engage with the AI community.
Referred to as a hub for AI development, encouraging founders to relocate or visit to gain insights and stay competitive.
Highlighted as a key location for AI innovation, with advice for founders to move there or at least visit to embed themselves in the ecosystem.
The video conferencing software whose frequent use during COVID led to the development of Superpowered, a tool to streamline joining calls.
Mentioned as a legacy software company that was disrupted by cloud-based alternatives, providing a parallel to current AI opportunities.
Mentioned as being only two years old, illustrating that even established AI models have significant room for improvement and that large companies are not necessarily faster executors.
More from Y Combinator
View all 562 summaries
14 minInside The Startup Reinventing The $6 Trillion Chemical Manufacturing Industry
1 minThis Is The Holy Grail Of AI
40 minIndia’s Fastest Growing AI Startup
1 minStartup School is coming to India! 🇮🇳
Found this useful? Build your knowledge library
Get AI-powered summaries of any YouTube video, podcast, or article in seconds. Save them to your personal pods and access them anytime.
Try Summify free