What we learned building agents at Vercel

Agents are increasingly being utilized by enterprises to boost productivity and improve outcomes, with applications in customer support, code reviews, and sales operations. The challenge lies in identifying suitable problems for AI solutions that provide value at a reasonable cost. Vercel has developed a methodology for selecting AI projects with the highest potential impact by focusing on tasks that require low cognitive load and high repetition, ideal for current AI capabilities. Two examples include a lead processing agent that automates research and qualification, significantly reducing manpower requirements, and an anti-abuse agent that cuts ticket processing time by 59% by automating the initial analysis of abuse reports. Vercel encourages organizations to explore AI applications by asking employees about repetitive tasks they dislike, offering templates for various agent applications, and providing hands-on support for discovering and implementing high-ROI AI projects.