OpenAI Presence: Why OpenAI Just Became Sierra's Biggest Rival

OpenAI launched Presence on July 22, 2026, a platform for deploying AI voice and chat agents that handle mission-critical business operations — customer support, sales development, procurement, IT help desks — with guardrails and human escalation built in. The product itself is a logical next step for OpenAI. The more interesting story is who it competes with: Sierra, the enterprise AI agent company run by Bret Taylor, who also happens to be OpenAI's board chairman.

There's no self-serve signup, no published API pricing, and no way to try Presence without going through OpenAI directly. That alone tells you who this is built for, and it isn't the market OpenAI has served with ChatGPT Work, its other 2026 agent launch. Here's what Presence actually does, how it compares, and why the Sierra angle matters more than the feature list.

Quick Take

• What launched: Presence, an enterprise platform for deploying AI voice and chat agents on "mission-critical" operational work — customer support, sales development, procurement, IT services, HR — with policy controls, guardrails, simulations, and a Codex-powered improvement loop.
• How you get it: Not self-serve. Every deployment goes through OpenAI's Forward Deployed Engineers or approved systems integrators, in a limited general-availability rollout. Pricing is negotiated per deployment; no public rate card exists.
• The proof point: OpenAI runs Presence on its own English-language phone support line, where it says the system resolves 75% of inbound calls without a human. OpenAI also says a Codex-powered feedback loop cut human handoffs by 15 percentage points within 10 days on that line.
• Who it competes with: Sierra and Decagon, the venture-backed startups that already sell this exact category of product to large enterprises — and Sierra is run by Bret Taylor, OpenAI's own board chairman.

What Presence Actually Does

According to OpenAI's announcement, Presence bundles together the pieces a company needs to put an AI agent in front of real customers: a place to define policies and standard operating procedures, guardrails that intervene automatically when a conversation drifts outside approved boundaries, a library of "approved actions" the agent is allowed to take, simulation tools to test it before launch, evaluation dashboards, and an improvement loop powered by Codex, OpenAI's coding agent, which reviews interactions and proposes behavior changes for a human to approve.

The guardrails example OpenAI gives is a retailer that restricts which product databases the agent can pull from when answering sales questions — the point being that Presence agents are meant to operate inside hard boundaries a business defines, not wander freely across whatever data they can reach. That's a meaningfully different design goal from a general-purpose assistant: it's built to be predictable and auditable at the expense of flexibility.

Deployment is the real signal here, though. There is no API key you can generate and start building against. Presence ships through what OpenAI is effectively running as a consulting arm — Forward Deployed Engineers embedded with the customer, or select global systems integrators — during this initial limited-GA phase. Pricing isn't published; it's scoped per contract based on use case and volume.

This Isn't ChatGPT Work in a Trench Coat

It would be easy to assume Presence is just a rebrand of ChatGPT Work, the multi-step task agent OpenAI shipped on July 9. It isn't. ChatGPT Work is aimed at your own employees — it plans research, builds spreadsheets, drafts documents, and works across your connected apps on an internal task. Presence is aimed at your customers: it's meant to sit on a support line or a sales chat widget and interact with people outside your company, under guardrails strict enough that a bank or airline would trust it with a live phone call.

That's a fundamentally different risk profile, which is why Presence comes with simulation testing, policy enforcement, and mandatory human-guided deployment instead of a self-serve toggle. OpenAI clearly sees these as two separate product lines solving two separate problems, not one agent wearing different skins.

The Part Most Coverage Undersold: Sierra

Presence lands in a market that already has a well-funded incumbent, and the incumbent's CEO sits on OpenAI's own board. Sierra, founded in 2023 by former Salesforce co-CEO Bret Taylor and ex-Google executive Clay Bavor, builds exactly this category of product — AI agents for customer-facing operations across voice, chat, and messaging. Sierra raised $950 million in May 2026 at a $15.8 billion valuation, counts more than 40% of the Fortune 50 as customers, and has passed $150 million in annual recurring revenue.

Taylor has been OpenAI's board chairman since the company's governance shakeup in late 2023, a role he holds simultaneously with running Sierra. The two roles aren't a legal conflict — Sierra builds on top of multiple foundation models, OpenAI's included, rather than competing to build its own — but Presence changes the picture. OpenAI is no longer just a model supplier Sierra depends on; it's now selling the same finished product, to the same enterprise buyers, through the same kind of white-glove, guardrail-heavy deployment Sierra built its business on. That's an unusual position for a company's own board chair to be in, and it's worth watching whether OpenAI's board addresses it explicitly.

Decagon, Sierra's closest independent competitor in enterprise conversational AI, is the other name buyers in this space typically evaluate alongside Sierra. Both compete on the same axis Presence now enters: reliability, guardrails, and measurable resolution rates rather than raw model benchmarks.

Should Your Business Care?

If you're a small or mid-size business, Presence isn't for you right now — there's no self-serve path, and OpenAI's own framing (Fortune 500-style deployments, Forward Deployed Engineers, negotiated contracts) points at large enterprises with the budget and internal teams to run a months-long implementation. For that segment, ChatGPT Work or a lighter automation layer via Zapier, Make, or n8n remains the realistic option.

If you're an enterprise buyer already evaluating Sierra or Decagon for customer support automation, Presence is now a legitimate third name on that shortlist — backed by OpenAI's own models and a live reference case (its support line) rather than a demo. The honest caveat: Presence is brand new, with no independent customer case studies yet, while Sierra and Decagon have years of production deployments and public results to point to. "Built by the company that makes GPT-5.6" is a credible pitch, not yet a proven one.

Frequently Asked Questions

Can I sign up for OpenAI Presence myself? No. There's no self-serve signup or public API. Deployments go through OpenAI's Forward Deployed Engineers or approved systems integrators, as part of a limited general-availability rollout.

How much does OpenAI Presence cost? OpenAI hasn't published pricing. Deployments are scoped and priced individually per customer contract, similar to an enterprise consulting engagement rather than a metered API product.

How is Presence different from ChatGPT Work? ChatGPT Work is an internal productivity agent for your own employees — research, documents, spreadsheets. Presence is built for customer-facing operations like support calls and sales chat, with stricter guardrails and mandatory guided deployment.

Does OpenAI competing with Sierra matter if I'm just choosing a vendor? Not directly for evaluation purposes — judge each platform on guardrails, reliability, and results. But it's a useful signal that this category is now contested by the largest AI labs, not just startups, which should mean faster iteration and more competitive pricing over time.

Is Presence proven to work well? OpenAI's strongest evidence so far is its own support line, where it reports a 75% resolution rate without human escalation. That's a real result but a single, self-reported case study — not yet the kind of independent, third-party verification that Sierra and Decagon have accumulated over several years in production.

Bottom Line

OpenAI Presence is a credible, well-built entry into enterprise customer-facing AI agents — but it launches as a consulting engagement, not a product you can try today, and it puts OpenAI in direct competition with Sierra, a company its own board chairman runs. For enterprise buyers already shopping this category, it's worth a look once independent results exist. For everyone else, the more usable OpenAI agent remains ChatGPT Work, covered in our full agent comparison. For more on how the major AI platforms are competing across products, see our Claude vs. ChatGPT vs. Gemini breakdown or browse our AI Tools coverage.