OpenAI has a new product called Presence, built for one job: running AI agents that actually talk to customers and handle real support tickets, not just internal chatbots for employees. It comes with a dashboard that tracks accuracy, call volume, and problem cases after the agent goes live. It also has a system where the agent's own conversations get reviewed and turned into proposed fixes for OpenAI's staff to approve.
The interesting part is not the agent itself. It is how OpenAI sells it. Presence is not a product you sign up for like ChatGPT.
There is no public price list and no self serve option. A company has to qualify, and the deal gets priced individually based on what that company needs. SoftBank is already an early user, running natural sounding Japanese language customer service conversations through it.
When a company's setup gets complicated, OpenAI sends in its own staff, called Forward Deployed Engineers, to sit inside the client's business and build the thing with them. That means connecting old systems, writing the rules the agent must follow, and running tests before launch.
This is not a new idea. Palantir built this playbook years ago: instead of selling software off the shelf, put your own engineers inside the client's building until the system works, then turn that one deployment into a repeatable product. It is widely seen as a major reason Palantir's stock recovered sharply after a rough start as a public company.
OpenAI is now copying that playbook directly, and so is Anthropic. In May, OpenAI backed the move with real money, launching a separate unit called the OpenAI Deployment Company with more than four billion dollars from investors including TPG. It also bought a consulting firm called Tomoro specifically to get about 150 of these engineers on staff from day one.
For a mid sized company reading this, the takeaway is simple: Presence is not for you yet. It is built for companies large enough to negotiate a custom contract and patient enough to have outside engineers embedded in their operations for weeks. The bigger signal is where OpenAI is headed: less a company that rents you access to a smart model, more a company that shows up, builds the thing, and charges enterprise consulting rates for it.
That shift matters for two more reasons. First, it puts OpenAI in direct competition with customer service software companies many businesses already use, like Salesforce, ServiceNow, and Zendesk, even while some of those same companies separately partner with OpenAI on other projects. Second, OpenAI has not said how Presence handles rules like the EU AI Act, which requires customer facing chatbots to tell users up front that they are talking to a machine.
Any business weighing a deal like this in Europe should get that answer before signing anything.