OpenAI Launches Presence: AI Agents With Engineers Attached
A customer reaching a business by phone - the channel voice agents are starting to cover. Stock photo (CC0).
On July 22, 2026 OpenAI announced Presence - a platform for enterprises to deploy voice and chat AI agents across customer support, outbound sales and internal IT. The most striking part is not the technology but the go-to-market: OpenAI sends its own engineers to deploy it on-site. It is not self-service. What does that tell Vietnamese businesses?
Presence launched July 22, 2026 and packages the full agent operating stack: policies and standard operating procedures, guardrails, approved actions, simulation testing, evaluation tools and a Codex-powered improvement loop. It already runs OpenAI's own English phone support line, resolving 75% of inbound calls with no human involved; the improvement loop cut human handoffs by 15 percentage points within 10 days. Presence ships through a limited program, deployed by OpenAI's Forward Deployed Engineers and select integrators, with no public pricing. The lesson for Vietnamese businesses: the value sits in deployment and operations, not just the model.
What did OpenAI announce?
Per VentureBeat and The Register, Presence is a dedicated enterprise agent platform supporting both voice and chat across three job families: customer support, outbound sales and internal IT service requests. Unlike raw API access, Presence bundles everything an operations team needs to run agents in production: policies and standard operating procedures (SOPs), guardrails that block out-of-scope behaviour, an approved-actions list, a simulation environment for pre-launch testing, evaluation tooling, and a post-deployment improvement loop powered by Codex. In short: OpenAI is not selling another model - it is selling the machinery that makes agents work inside a company.
Figures OpenAI announced at the Presence launch on July 22, 2026, measured on its own English support line. Self-reported numbers - treat as indicative.
The telling detail: not self-service
Presence ships through a limited general availability program, deployed on-site by OpenAI's own Forward Deployed Engineers together with a handful of selected systems integrators. The Register describes it as OpenAI "trying the consulting path" at boots-on-the-ground rates - cost is negotiated per customer and per deployment, and no public price sheet exists. It echoes the message of the PwC-OpenAI customer service partnership two weeks earlier: the biggest players all concede that agents are not install-and-run software, but deployment projects with humans attached.
Agents run well when people tune each script against real conversations. Stock photo (CC0).
The Chạm AI view: lessons for Vietnamese businesses
First, Presence confirms what agent builders have long argued: the model is only one part - the value sits in the operating stack of procedures, guardrails, simulation testing and continuous improvement on real conversations. Those are exactly the steps in our own AI agent implementation process. Second, even OpenAI - with the strongest models - sends engineers on-site, which means no business should expect to buy a licence and be done. Third, Presence targets large enterprises at international consulting rates; Vietnamese SMEs are almost certainly not the intended customer, but they can absolutely ask a local partner to build the same "Presence stack" - SOPs, guardrails, testing, an improvement loop - at Vietnamese cost, on Vietnamese channels such as Zalo OA or Messenger. Whoever standardises operations wins, regardless of the model underneath.
Sources: VentureBeat - "OpenAI unveils Presence" · The Register - "OpenAI tries the consulting path with Presence", July 22, 2026 · Help Net Security, July 22, 2026. The 75% and 15-percentage-point figures are OpenAI's own, measured on its own support line.
Frequently asked questions
What is OpenAI Presence?
Presence is the platform OpenAI announced on July 22, 2026 for enterprises to deploy voice and chat AI agents across customer support, outbound sales and internal IT. It packages the full operating stack: policies and standard operating procedures, guardrails, approved actions, simulation testing, evaluation tools and a Codex-powered improvement loop.
Is Presence generally available, and what does it cost?
Not yet. Presence launches through a limited availability program, deployed on-site by OpenAI's own Forward Deployed Engineers and select systems integrators - it is not self-service. OpenAI has not disclosed pricing; cost is set per customer, and observers describe it as enterprise consulting rates.
What should Vietnamese businesses take away from Presence?
Three things: production agents need a full operating stack (procedures, guardrails, testing, continuous improvement), not just a model; even OpenAI sends engineers on-site, which means deployment is the hard part; and since Presence targets large enterprises at enterprise rates, Vietnamese SMEs should ask local partners to build the same stack at local cost on local channels.