Retool Weds GPT-4 to Business Apps, Touts Trillions in Stuck Value

The gap between custom software and a spreadsheet is where Retool has always played. Now the company is shoving GPT-4 into that gap and betting the combination can unstick trillions of dollars in business value currently trapped in manual processes. Announced this month, Retool AI is a suite of pre-fab building blocks—text generation, image description, and other common AI actions—dropped directly onto the platform’s visual canvas. The whole thing runs exclusively on OpenAI’s models.

Cory Wilkerson, Retool’s engineering lead, described the shift in plain terms: apps that used to be static and transactional now get ‘activated’ with a layer of intelligence that wasn’t accessible before. That’s not marketing froth. It’s a real architectural change for internal tools. Instead of building rigid forms and workflows, developers can now wire in a component that understands freeform questions and reasons over company documentation.

Chief Technology Officer Anthony Guo pointed to inventory management, customer support, and tailored product experiences as early use cases where customers are already seeing ‘huge cost savings.’ The support chatbot is the obvious starting gun. GPT-4 can ingest a company’s website and docs, then field questions from employees or customers without someone first mapping every possible query to a decision tree. Retool’s pitch is that you get this capability without sacrificing data governance—the platform handles access controls and security practices that a bespoke integration might skip.

The roadmap includes expanding the library of pre-built AI Actions and shipping dozens more app templates. It’s a land grab for the massive middle of enterprise software: processes too specific for off-the-shelf SaaS but too costly to build from scratch. Retool is essentially saying you can now have both the custom fit and the AI smarts, assembled on a canvas rather than architected over six months. Whether GPT-4’s occasional hallucinations play nice with business-critical operations remains the uncomfortable question nobody’s fully answered.