Enterprise leaders have joined a reliable program for nearly two decades. VB transform brings people to develop real enterprise AI strategies. Get more information
Scott White is still surprised how artificial intelligence has changed a real work partner from a novelty. Just a year ago, the product leads Claude o at all Anthropic Early AI coding tools can barely complete the same line of code. Today, he is producing a production software despite not being a professional programmer.
“I no longer try to write a PRD about my job and to please anyone to do something,” White said during a fireide chat. ” VB Transform 2025Venture Bat’s annual Enterprise AI Summit in San Francisco. “I have done this first, can I make a viable prototype on my staging server and then share a demo that is actually working.”
This change represents a broader change of how businesses are adopting AI, and moving beyond easy chat boats that answer questions for sophisticated “agent” system. The White experience offers a glimpse of what can happen for millions of other academic workers.
From the completion of the code to the autonomous programming: AI breakdown evolution
Evolution has been significantly sharp. When the White Anthropic joined the company Claude 2 The model can handle the completion of the basic text. Release of Claude 3.5 Swant Enabled the creation of entire applications, which features Sampling Let users prepare customs interfaces. Now, with Claude 4 Getting 72.5 % score on SWE-BENCH coding benchmarkThe model can work in a way that is called White “Fully remote agent software engineer.”
Claude codeThe company’s latest coding tool, can analyze the entire code base, can find the Internet for API documents, release bridges requests, respond to code review comments, and repeat the solutions – all of these work seriously for hours. White noted that 90 % of the cloud code was written by the AI ​​system itself.
White explained, “This is like a whole agent process in the background that was not possible six months ago.”
Enterprise Giants Slush Work Time with AI agents with AI agents
Its implications are much more than software development. Novo NordeskDanish Pharmaceutical Dev, has merged the cloud into workflows, before it took 10 weeks to complete the clinical reports, which now ends the same work in 10 minutes. The gut lab Uses technology for everything from sales tips to technical documents. The Antote deployed the cloud to provide direct tax advice to consumers.
There is a difference between different levels of white AI integration: the simplest models of the language that answer questions, models are enhanced with web search, such as tools, structured workflows that include AI in the business process, and the full agent that is using numerous tools and repetitive reasoning.
“I think of an agent who has a goal, and then can do a lot of things to fulfill that goal,” said White. The key personnel has been the one who calls the “unmanageable” relationship between model intelligence and new product capabilities.
Infrastructure Revolution: AI’s colleagues’ building networks
Has been a significant development of infrastructure Anthropic model context protocol (MCP), which White describes as “USB-C” for integration. Instead of creating separate contact companies from each data source or tool, MCP provides a standard method for the AI ​​system for access to enterprise software from sales force to internal academic deposits.
“This is really democratic access to data,” said White, saying that the integration built by a company can be shared and reused by others by open source protocol.
For organizations wanting to enforce AI agents, White recommends increasing small and construction. He advised that “don’t try to create a whole agent system from the beginning.” Make it a component, make sure the component works, then make the next component. “
He also emphasized the importance of the diagnostic system to ensure AI agents to perform the purpose. “Eules are the new PRD, referring to the product requirement documents, while highlighting how companies have to develop new ways to evaluate AI’s performance on specific business work,” White said.
From AI Assistants to AI organizations: Next workforce Frontier
Looking forward, the White imagines that the development of AI becomes accessible to non -technical workers, how the same coding capabilities have been developed. He considers a future where individuals manage not only an AI agent but the entire AI system organizations.
“How can everyone have their own mini -CPO or CEO?” White asked. “I don’t know exactly what it looks like, but it’s something I wake up and want to get there.”
Change White states that the industry’s wider trends are reflected as companies suffer from AI expansion capabilities. Although the initial lap is focused on experimental use issues, businesses are rapidly integrating AI into basic business processes, changing how basically works.
Since the AI ​​agents are more independent and capable, the challenge of the AI ​​is converted to the Teaching Challenge teaching machines to handle the task of handling the work that can work freely for extended periods. For white, this future is already coming – a production feature at a time.
