OpenAI announced OpenAI Presence on July 22, 2026. It is a managed product for deploying enterprise AI agents across voice and chat workflows such as customer support, outbound sales, billing, insurance claims, and employee IT requests.
For marketers, the important part is not another model release. Presence packages the difficult operating layer around agents: company data access, policies, approved actions, testing, escalation, measurement, and controlled updates after launch. That moves the conversation from “Can an AI answer?” to “Can it complete a real customer task without creating a new support problem?”
What was announced
OpenAI Presence starts with a narrowly defined job. The agent receives the knowledge and system access required for that job, while the company sets rules for what it can do, when approval is required, and when a person must take over.
The product combines standard operating procedures, guardrails, simulations, evaluation tools, approved actions, and a Codex-powered improvement process. After deployment, teams can review production sessions and escalations. Codex can then suggest changes, which teams test and approve before rollout.
Presence currently supports real-time voice and chat. OpenAI says it uses the system for its English-language phone support line, where the agent can verify callers, use account context, resolve eligible billing issues, and escalate harder cases.
OpenAI reports that Presence resolves 75% of inbound issues on that support line without human help. It also says a Codex-powered improvement loop reduced human handoffs by 15 percentage points in 10 days. Those are vendor-reported results from OpenAI’s own deployment, not independent performance benchmarks.
OpenAI also named three enterprises building on Presence. BBVA is exploring AI-powered voice support for everyday banking in Mexico, SoftBank is testing Japanese-language customer conversations, and IAG is exploring support during high-demand periods such as severe weather. All three are described as exploratory or in testing rather than completed rollouts.
Availability is limited. OpenAI Presence is offered to eligible enterprise customers through a limited general availability programme. Deployments are led by OpenAI Forward Deployed Engineers and selected systems integrators. It is not a self-serve product, and broader pricing has not been published.
Why OpenAI Presence matters for marketers
The clearest marketing impact sits at the handoff between acquisition and service. Most growth teams spend heavily to create demand, then pass prospects into forms, SDR queues, demo scheduling tools, support inboxes, and CRM workflows that do not share enough context. Presence is aimed at the operational gap between the message that won the click and the action the customer needs next.
Six things change for marketing teams.
- The voice layer finally has operating controls. OneMetrik’s earlier coverage of GPT-Live examined how real-time conversation could affect research, onboarding, and support. Presence adds the controls required to turn that capability into a business process.
- Voice AI agents stop being a support-only concern. A qualified prospect could ask a detailed pricing question, verify eligibility, book the correct sales route, or receive an approved follow-up without waiting for a rep to copy information across three systems.
- Response speed matters more than outreach volume. This is where AI automation in sales becomes more useful than generic email generation. The value is shortening response time, keeping account context intact, and routing the buyer correctly.
- Customer support automation becomes brand execution. An agent may be technically correct and still damage trust if it sounds evasive, applies policy inconsistently, or fails to recognise when a person should step in. Marketers will need input into tone, claims, escalation rules, offer language, and the way the agent explains product limitations.
- B2B SaaS marketing picks up a new content requirement. Product pages, help documentation, pricing rules, objection handling, and sales enablement material must be clear enough for an agent to use. Vague positioning does not become better when spoken by a faster system.
- Workspace agents solve a different problem. Teams already testing ChatGPT workspace agents will recognise part of the model, but those focus on repeatable team workflows inside tools such as ChatGPT and Slack. Presence is aimed at production interactions with customers and employees, with deployment support built around each use case.
How OpenAI Presence compares with alternatives
OpenAI Presence is not the cheapest or fastest route for every company. Its main difference is the managed deployment model.
A custom OpenAI API build gives a company more control over architecture and vendor choices, but the company must build its own policies, monitoring, testing, approvals, integrations, and maintenance process. OpenAI says it will continue supporting voice customers through its API alongside Presence.
OpenAI workspace agents are better matched to shared internal work such as reporting, research, and handoffs. Presence is positioned for live voice and chat interactions where the agent may verify identity, use customer data, or take an approved action.
The Grok Voice Agent Builder takes a more self-serve route. xAI describes it as a no-code beta for configuring production voice agents with telephony, knowledge retrieval, tools, guardrails, and monitoring, priced at $0.05 per minute of agent audio. That may suit faster experiments, while Presence is designed around individually scoped enterprise deployments and has no published pricing.
Traditional chatbot and contact-centre platforms remain another option. They may be easier to buy and operate for predictable flows, but they can be less flexible when a request requires reasoning across policy, account context, and several connected systems.
The right comparison is not which agent sounds most human. It is which operating model matches the risk, data access, team capacity, and cost of failure.
What marketing teams should watch next
Start with one workflow, not a promise to automate the entire funnel.
A sensible first test has a clear beginning, approved actions, measurable outcomes, and an obvious escalation point. Lead qualification, demo routing, renewal questions, basic billing support, or event follow-up are easier to evaluate than an open-ended “AI sales agent.”
Before testing, define five numbers: containment rate, escalation accuracy, resolution time, conversion to the next step, and human review cost. Our guide to AI marketing ROI explains why time saved means little if pipeline, service quality, or revenue does not improve.
Marketing teams should also watch how Presence feeds interaction data back into campaign decisions. Voice and chat sessions can expose objections, pricing confusion, product gaps, and message failures faster than quarterly surveys. That data could improve a SaaS GTM strategy, but only if it reaches content, paid media, product marketing, sales, and customer success.
The governance question is equally important. Any AI marketing automation connected to customer records, billing, or CRM actions needs limited permissions, clear approval rules, logged decisions, and human escalation. The wider AI marketing automation stack should reduce repetitive work without hiding who is accountable when the system gets something wrong.
OneMetrik Takeaway
OpenAI Presence is a signal that enterprise AI agents are becoming an operating system problem, not a prompt-writing problem.
The companies that benefit will not be the ones with the most impressive demo. They will be the ones with clean policies, usable data, narrow workflows, strong evaluation, and a clear answer to one basic question: what happens when the agent is wrong?
For marketers, the practical move is to map one high-friction customer or lead handoff, document the approved actions and escalation rules, then measure whether an agent improves the next business outcome. Presence may be built for large enterprises today, but the operating discipline behind it applies to any team using voice AI agents, sales automation, or customer support automation.