OpenAI has unveiled Presence, a new enterprise platform for deploying AI agents that is already available for voice and chat systems. It is not just another model but a complete package: systems, evaluation tooling and expertise for adapting agents as business conditions change. OpenAI is betting that enterprises no longer need proof that AI agents work — what they need is reliability for high-value tasks in production.
What it is
OpenAI has launched Presence, a product aimed at enterprise customers that want to embed AI agents into their workflows. The platform lets agents answer questions, resolve issues, use companies' internal systems, take approved actions and escalate requests to people when needed. Presence has already proven itself at leading enterprises, including OpenAI's own internal support desk, where it handles 75% of incoming requests without a human and cut handoffs to people by 15 percentage points in 10 days. Other companies using Presence include BBVA for voice support in Mexico, SoftBank for Japanese-language customer communications and IAG for support during peak load.
Every Presence rollout starts with a specific task — resolving billing problems or processing insurance claims, for instance. The agent gets only the knowledge and system access it needs. Companies set the policies: what the agent may do, when approval is required and when a human must step in. Once live, data from real sessions and escalations surfaces the gaps, and Codex proposes updates that teams can test and approve, helping the agent adapt as customer behaviour changes.
Why it matters
The launch of Presence signals that the enterprise AI-agent market has matured. This is not a capability demo but a finished, battle-tested product that solves the key problem: making agents reliable and controllable inside real business processes. The emphasis on policies, safety rules and escalation paths shows OpenAI understands how critical control is in an enterprise setting. It changes how AI gets adopted — from experiments to strategic deployment.
Presence also underlines how much post-launch adaptation matters. Unlike static models, Presence agents can learn and improve from real interactions, which is critical in a fast-moving business environment. That makes investment in AI agents a long-term commitment rather than a one-off, since the agents evolve alongside the company and its needs. For developers it means moving from writing throwaway scripts to designing durable, evolving systems.
What it means in practice
- Focus on architecture and control: if you build agents for business, pay attention not only to the model but to management systems, policies, escalation rules and audit mechanisms. Those are the foundations of trust and scale.
- Iterative improvement: build feedback loops and automated processes into your projects so agents keep improving on real data. Agents have to adapt to changes in products, policies and user behaviour.
- Integration with existing systems: design agents for seamless integration with corporate knowledge bases, CRM, ERP and other internal systems. That is what makes them valuable and functional.
- Hybrid solutions: don't be afraid to mix automation with human involvement. Presence shows agents can work effectively, but there must always be a path to escalate to a human for complex or high-risk situations.
- Testing and simulations: before pushing agents to production, run thorough testing and simulations to surface potential problems and confirm the agent matches your policies and quality standards.