OpenAI Presence: new platform to build and manage business AI agents

Full text

OpenAI Presence is a new tool to monitor and update AI agent behavior

It works across voice and chat agents from launch, including OpenAI's phone support

Presence is led by forward-deployed engineers who already know an organization inside-out

OpenAI has launched a new, dedicated platform for building, deploying and iterating AI agents for business workflows, which it's calling OpenAI Presence.

From launch, the tool focuses on voice and chat agents that can answer questions, resolve problems, access company systems, perform approved actions and liaise with humans.

The addition comes in response to what the company sees as the challenge of keeping agents reliable after deployment.

Presence keeps your agents current and relevant

For example, OpenAI explained how agent behavior may need to change as company policies, products, systems and customers evolve.

With agents only granted access to the knowledge and systems needed for a particular task, the ChatGPT maker wants to keep decisions like what the agent is authorized to do, which actions require human approval, which policies must be followed and when the interaction must be escalated to company admins.

"Every deployment generated insights that continuously inform OpenAI research and product development, creating compounding benefits for customers," the company noted, adding that Presence comes from years of first-hand experience dealing with customers.

Presence also includes a Codex-powered improvement process designed to help agents continue evolving once they've been launched, with the agentic coding assistant capable of investigating behavior and proposing changes.

OpenAI also used its announcement to plug its forward-deployed engineer (FDE) program, which inserts OpenAI experts within customer organizations to build AI tools and processes tailored to their exact needs.

The new tool already powers the company's English phone support service (1-888-GPT-0090), handling open-ended requests, accessing account information and taking actions across billing and refunds.

Presence deployments are also FDE-led, and certain eligible enterprise customers can get access to the scheme through a limited general availability program.