Emergence is building AI agent infrastructure for complex enterprise workflows, drawing on experience with large-scale systems such as IBM Watson and Alexa.
2
Its Orchestrator is a meta-agent that plans, acts, verifies, remembers, and routes work across multiple agents and language models.
3
Agent E is an open-source web agent designed to operate websites like a person and automate workflows such as making reservations and sending the details to a colleague.
Summary
Dr. Satya V. Nitta introduces Emergence as an R&D company focused on the science and development of AI agents. The company combines distributed-systems experience with work from IBM Research, Google Brain, Alexa, Meta, and Microsoft. Nitta argues that AI is finally reaching the point where it can take actions in browsers and enterprise software, where workflows are more demanding than simple consumer tasks. Emergence is developing an Orchestrator that plans, acts, verifies, remembers, and routes work across different agents and language models. It is also integrating the Orchestrator with Agent E, an open-source web agent that learns to operate the web and automate enterprise workflows. The company's research focuses on self-improvement, planning, reasoning, and Agent Oriented Programming, with examples including RPA and document processing.
Emergence combines AI research with experience running large distributed systems
Nitta describes Emergence as an R&D AI company whose team comes from major AI labs, including IBM Research, Google Brain, Alexa, Meta, and Microsoft. He also points to deployments such as IBM Watson, Alexa, and recommendation systems behind Amazon Prime, Amazon Video, and Twitter. The company thinks of itself as a distributed-systems and AI research team. Its goal is to help developers build agents that can take useful actions at scale.
Enterprise software is where agent systems face their hardest workflows
Nitta says the long-standing promise of AI has been that software will perform actions for people. Emergence is focused on agents that operate web browsers, enterprise systems, and other software. He expects productivity gains in both consumer and enterprise settings, but chooses enterprise because its workflows are more interesting and demanding. These workflows will push the limits of autonomous AI, especially when an agent must complete several connected steps rather than answer a single request.
The Orchestrator routes multiple agents through complex workflows
Emergence's first infrastructure platform is an Orchestrator agent, with early access already available and a broader launch planned for August. Nitta describes it as an agent that acts, plans, verifies, remembers, and improves over time. It can coordinate several agents and connect them into a workflow. A basic use could route tasks across commercial and open-source language models. More complex uses include enterprise processes such as claims processing.
An orchestrator depends on the quality of the agents it coordinates
Nitta says an orchestrator is only as good as the agents it can orchestrate. Emergence is integrating its Orchestrator with Agent E, an open-source project developed in the company's R&D labs. Agent E is a web agent intended to control websites in a human-like way. Nitta describes it as a foundation for enterprise workflows and says it is designed to work with the Orchestrator rather than operate as an isolated web automation tool.
The product combines routing, model selection, and developer controls
The product demonstration describes a system that combines an Orchestrator with task-specific language models and agents, using guardrails for the different tasks. Its developer dashboard is intended to help teams analyze prompts, adjust cost and latency, create models with their own data, and improve existing models. The system is also designed to move to newer language models when needed, so applications do not have to be rebuilt around one fixed model.
Agent E is presented as an open-source web agent that learns to operate the web autonomously and automate complex workflows. The example request is to find a reservation for 15 people near the office at around 6 p.m., within a set of cuisine preferences, then send the details to Levi at Emergence. The example shows the kind of task Emergence wants agents to complete across websites and communication steps.
Emergence's research focuses on self-improvement and agent composition
Nitta closes by describing self-improvement agents as the company's central research focus. The work also covers AI planning and reasoning, along with ways to connect agents into useful systems through Agent Oriented Programming. Emergence is applying this research to enterprise workflows, including robotic process automation and document processing. The emphasis is on making agents improve and cooperate across longer business processes.
"Agent E is an open source agent designed to learn how to autonomously operate the web and automate complex workflows."04:24
Who should watch
You are designing enterprise software that needs to coordinate several AI agents or language models across one workflow.
You are evaluating browser automation for tasks that involve multiple websites, forms, and follow-up actions.
You want to understand Emergence's approach to agent planning, verification, self-improvement, and model migration.
