CrewAI today made generally available an enterprise edition of a platform that makes it possible to build and orchestrate multiple artificial intelligence (AI) agents using any type of large language model (LLM).
Fresh off raising a total of $18 million in funding, CrewAI CEO João Moura says the enterprise edition of a previously launched open source edition of the platform adds templates that make it simple to build agents along with additional built-in access controls and customer support services. “It’s based on an open source library,” he adds.
The company claims the CrewAI platform is already being used to manage more than 10 million agents per month and is used by nearly half of the Fortune 500 list of organizations. In total, CrewAI signed up 150 organizations as beta customers in less than six months for the enterprise edition, and that it is tracking 100,000 groups of multi-agent executions per day across hundreds of different use cases.
Available both as software-as-a-service (SaaS) and self-hosted editions, the CrewAI platform makes it simpler to create custom AI agents that are assigned specific roles. It currently supports sequential task execution and hierarchical processes, but the ability to manage more complex consensual and autonomous processes will be added. A Crew Studio tool makes it simpler to build those complex interactions.
In the meantime, AI agents can autonomously delegate tasks and interact with each other to resolve issues. The platform also enables self-iteration, performance evaluation, a wide range of collaboration structures and the ability to provide AI agents with access to persistent memory.
AI agents are providing organizations with a more flexible alternative to robotic process automation (RPA) platforms that are too brittle to be used in, for example, marketing workflows that require an ability to reason, notes Moura.
It’s not clear how rapidly organizations are adopting frameworks to build and manage AI agents, but the one thing that is certain is multiple AI agents can be used to manage increasingly complex tasks. The issue now is determining which tasks lend themselves to be performed by AI agents that are then supervised by humans.
While there is little doubt at this point that the rise of AI agents will disrupt the current level of demand for workers that previously performed many of those tasks, the fact remains that many of the tasks that are being automated using AI agents are not ones that anyone especially enjoyed. The overall goal is to enable organizations to operate at a level of scale that would previously have been impossible to achieve, either because there were simply not enough human workers available or the cost of hiring them would have been prohibitive.
There will, of course, always be a need to keep humans in the workflow loop to ensure AI agents are performing tasks as expected. In addition, there will always be some exception to every rule that will require some level of human intervention to manage. AI agents, however, are already here to stay, and the next challenge is determining precisely to what degree they can be relied on to perform the tasks assigned to them.
