From Chatbots to collaborators: How AI agents reshape the work of the institution

Photo of author

By [email protected]


Join the event that the leaders of the institutions have been trusted for nearly two decades. VB Transform combines people who build AI’s strategy for real institutions. Learn more


Scott White is still surprised by the rapid transformation of artificial intelligence from modernity to a real business partner. A little more than a year ago, the product leads to Claude AI in man I watched artificial intelligence coding tools early can hardly complete one line of code. Today, he is building ready-to-produce programs for himself-although he is not a professional programmer.

White said during a conversation on the fire side: “I no longer think about my job as writing PRD and I try to persuade someone to do something,” White said during a conversation on the fire. VB converting 2025Venturebeat annual institution summit in San Francisco. “The first thing I do is, can I build a first business model for this on our graduate server and then share a trial display of it already.”

This transformation represents a broader shift in how institutions are adopted by Amnesty International, as they exceed the simple chat groups that answer questions on advanced “agent” systems capable of independent work. White experience provides a glimpse of what may come to millions of other knowledge workers.

From completing the code to independent programming: Breakneck’s development of artificial intelligence

The development was significantly rapid. When White joined Anthropor, the company Claude 2 The model can deal with the completion of the basic text. release Claude 3.5 Sonata It was possible to create entire applications, which leads to features such as Artifacts This allows users to create custom facades. Now, with Claude 4 Achieve 72.5 % on Swe-BenchThe model can work as White calls it “a completely distant agent software engineer.”

Claude iconThe latest coding tool for the company, the entire Codebases analysis, can search on the Internet for API documents, issue withdrawal requests, respond to the comments review of code, and repeat solutions – all of this while working as an unsafe for hours. White note that 90 % of the Claude code itself was written by the artificial intelligence system.

“This is like an entire agent in the background that was not possible six months ago,” White explained.

The Giants of the Foundation cut the time of work from weeks to minutes with artificial intelligence agents

The effects of the software development. Novo NordskThe Danish drug giant, combined Claude into the work progress that previously took 10 weeks to complete clinical reports, which now end the same work in 10 minutes. Gitllab Technology is used for everything from sales proposals to technical documents. Intuit Claud is published to provide tax advice directly to consumers.

White distinguishes between different levels of artificial intelligence integration: simple language models that answer questions, models reinforced with tools such as searching on the web, the functioning of organized work that integrates artificial intelligence into commercial processes, and the full agents that can follow the goals independently using multiple tools and repetitions.

White said: “I think about an agent as something that has a goal, and then he can do many things to achieve this goal,” White said. The main ancient factor was what the “invincible” relationship between model intelligence and the capabilities of new products.

Infrastructure Revolution: Building networks of collaborators with artificial intelligence

The development of critical infrastructure was The typical context protocol of anthropology (MCP), which White describes as “USB-C for Integration”. Instead of companies that build separate contacts for every source or data tool, MCP provides a unified method of artificial intelligence systems to access institutions programs, from Salesforce to internal knowledge warehouses.

“It is really democratic access to data,” White said, noting that the integrations designed by one company can be shared and reused by others through the open source protocol.

For organizations looking to implement artificial intelligence agents, White recommends a small start and a gradual construction. “Do not try to build a complete agent system from scratch,” advised. “Build the component of it, make sure the component works, then create the next component.”

He also stressed the importance of evaluation systems to ensure the performance of artificial intelligence agents in an intended manner. “Evals is the new PRD,” White said, referring to the product requirements documents, while highlighting how companies develop new ways to assess the performance of artificial intelligence in the specified work tasks.

From artificial intelligence assistants to artificial intelligence organizations: the next workforce is the workforce

Looking at the future, White imagines the development of artificial intelligence for non -technical workers, similar to how coding capabilities developed. He imagines a future in which individuals do not manage only one customer, but complete organizations of specialized artificial intelligence systems.

“How can everyone be a miniature CPO or CEO?” He asked White. “I don’t know exactly what it looks, but this is the thing I wake up and want to get there.”

The Transformation White reflects the broader industry trends as companies wrestle with the expanding capabilities of AI. Although early adoption focuses on experimental use cases, institutions are increasingly integrating artificial intelligence into basic commercial processes, which mainly changes how to do work.

When artificial intelligence agents become more independent and capable, the challenge from teaching machines turns into tasks to manage collaborators with artificial intelligence who can work independently for long periods. For eggs, this future already reaches – one production feature simultaneously.




Source link

Leave a Comment