How To Use AI In Your Startup

The panel advises founders not to pivot to AI just because it is popular, while arguing that new companies should look for ways to use LLMs. One example is an HOA management company that uses LLMs internally to make its work more efficient. Rebuilding existing software with AI at its core may also create an opening, much like cloud software did for on-premise products.

The panel compares the current AI wave with earlier shifts such as mobile, when new capabilities opened opportunities for startups. A story about Superpowered describes founders who first built a Zoom productivity tool, then stopped working on it and pursued a new company as LLMs advanced. The panel cautions that a weak business will not improve just by changing its idea to one that uses OpenAI; founders still need customer insight and a sound approach.

The panel recommends spending time around people building with AI to understand what is possible and useful. One speaker recalls Airbnb asking Pinterest, located nearby, for help learning SEO. For founders who cannot move to the Bay Area, the panel suggests visiting for three or four weeks, attending hackathons, and meeting people in the startup community.

Examples from a recent YC batch include Replex, which automates UI translation, and Gecko Security, which acts as an AI security engineer. Both put specialized work behind an interface that lets software engineers manage it themselves. Another company used prior experience in insurance technology to build an AI copilot for Medicare Advantage agents.

The panel describes healthcare administration as a mix of legacy software and people moving information between systems. Tara uses LLMs to prepare prior authorization requests from doctors’ information, while another company uses voice AI to check on patients between visits and help schedule follow-ups. The panel suggests observing healthcare workers’ screens to spot repetitive tasks that could be improved.