OpenAI continues to expand its enterprise offering around agents, applied research and adoption. The centrepiece of this phase is OpenAI Presence, introduced on July 22: a managed product for deploying voice and chat agents in specific customer-service and internal-operations workflows.
Presence is neither a general-purpose chatbot nor a self-service product. Each deployment starts with a narrowly defined job, receives only the knowledge and access it needs, and operates under policies set by the organisation.
What Presence does
- Connects the agent to business systems and pre-approved actions.
- Lets organisations define permissions, policies, evaluations and escalation rules.
- Tests common requests, edge cases and higher-risk scenarios before launch.
- Reviews sessions and escalations after deployment to identify failures.
- Uses Codex to propose improvements that teams can test and approve before rollout.
OpenAI says Presence already powers its English-language phone support and resolves 75% of inbound issues without human assistance. That is an internal OpenAI metric, not a performance guarantee for every company.
Enterprises and news organisations
OpenAI names BBVA, SoftBank and IAG among the organisations exploring the platform. Presence is offered through a limited general availability programme, with deployments led by OpenAI engineers and selected partners; it is not yet self-service.
The enterprise push also reaches the media industry. On July 22, OpenAI published examples of news organisations using AI to review overnight information, structure documents, explore archives and prepare signals that are then checked by human teams.
What changed on July 29
OpenAI's news page lists three new publications that day: a technical explanation of how GPT‑5.6 combines capability and efficiency, a programme for accelerating scientific research with ChatGPT, and an analysis of ARC‑AGI‑3.
It does not list another flagship model released after GPT‑5.6. The day's signal is therefore not a new model name, but the conversion of existing models into enterprise products, workflows and infrastructure.
Why it matters
AI competition no longer depends only on who scores best on a benchmark. Enterprises also care about system integration, permission controls, operating cost, action traceability and the ability to hand an exception to a person.
Presence moves OpenAI into that operational layer. If the approach works, the advantage will lie less in an isolated demo and more in keeping agents dependable as policies, products and user behaviour change.
What remains unclear
- Pricing and commercial terms.
- The timetable for broader availability.
- Independent quality, cost and error-rate measurements.
- Which integrators can deploy Presence in each market.
- When channels beyond voice and chat will become available.
Did OpenAI launch another flagship model on July 29?
No. It published GPT‑5.6, scientific-research and evaluation updates, but did not announce a successor.
Is Presence available to everyone?
No. OpenAI offers it to eligible enterprise customers through a limited general availability programme with assisted deployments.
Can Presence act without human control?
Only within approved permissions and actions. The organisation defines when it must request approval or escalate to a person.
Verified sources and original reporting.
Nexus AI wrote and contextualized this article using OpenAI · Presence and News · Jul 22–29, 2026. The complete story is on this page; the reference is provided so readers can check the original information.
Check the main source ↗
