OpenAI has launched Presence, a production-grade platform that pairs its AI agents with forward-deployed engineers to help enterprises turn experimental bots into reliable, mission-critical services. The move signals the company’s intent to become a core infrastructure provider for businesses that need AI to run at scale, not just as a curiosity lab.

From internal GPTs to enterprise-ready agents

For the past year OpenAI’s main enterprise offering has been Workspace Agents – customizable “GPTs” that companies use for internal productivity tasks. Those agents handle simple queries and document generation, but they crumble under high-stakes, customer-facing applications. Presence fills that gap. It targets automated customer support and complex workflow orchestration, where downtime or unexpected behavior can directly hit revenue and brand reputation.

The platform’s architecture diverges from a standard GPT deployment. Instead of handing a client a black-box model and a set of prompts, OpenAI bundles the model with engineering services that stitch the agent into existing IT ecosystems. The goal is to move AI from a helpful assistant to a dependable component of a company’s operational backbone.

Forward-deployed engineers: the human bridge

OpenAI’s most visible differentiator is the deployment of forward-deployed engineers (FDEs) who work side-by-side with qualifying customers. These engineers become part of the client’s implementation team and handle four core responsibilities:

  • Workflow architecture – Map business processes to AI capabilities and pick the most effective interaction patterns.
  • Systems integration – Hook the agent up to legacy software, proprietary databases, and other in-house tools.
  • Guardrail implementation – Define strict behavioral limits to keep the agent’s output within acceptable bounds, reducing hallucinations or policy violations.
  • Lifecycle management – Run rigorous testing cycles, move the agent from sandbox to live production, and monitor performance over time.

By embedding engineers in the deployment, OpenAI aims to smash the “plug-and-play” myth that many AI vendors tout. The engineers shape the model’s raw capability with the client’s data, security requirements, and compliance constraints.

The compliance and transparency hurdle

Enterprise adoption of AI remains constrained by regulatory and governance concerns. Presence is offered only to qualifying enterprises, and OpenAI has not released detailed compliance documentation. The EU AI Act, for example, imposes strict rules on high-risk AI systems, but OpenAI’s public statements stop short of describing how Presence meets those obligations.

OpenAI cites “trust mechanisms” – audit logs, model explainability tools, and data-privacy safeguards – yet the exact legal frameworks stay vague. For developers and founders, the tension is clear: AI innovation outpaces the development of globally accepted standards. Companies that adopt Presence now must weigh early-access benefits against the risk of future regulatory retrofits.

Who wins, who watches, and what it costs

OpenAI stands to gain a foothold in the lucrative enterprise-software market, where reliability and support contracts often dwarf consumer-grade licensing fees. Forward-deployed engineers generate recurring revenue and lock customers into long-term relationships.

Enterprises that integrate Presence could see faster automation of repetitive tasks, fewer support tickets, and more consistent customer experiences. However, the service’s cost structure – model usage fees plus engineering labor – remains undisclosed, leaving smaller firms to wonder if the price fits their budgets.

Competitors that have historically offered self-serve AI APIs may feel pressure to add similar professional services, or risk losing high-value accounts that can’t afford the trial-and-error of in-house integration.

Counterpoint: the risk of vendor lock-in

Critics argue that bundling engineering talent with the AI model creates vendor lock-in. Once a company’s core processes revolve around OpenAI’s agents and custom guardrails, migrating to a different provider could become costly and technically complex. The lack of an open, standards-based interface for the guardrails tightens that grip.

Presenceの利用が「要件を満たす」顧客のみに限定されていることは、多くの中堅企業が初期の波に取り残される可能性があることを意味しており、AIのリーダーと遅れをとる企業との格差を広げる恐れがあります。

今後の注目点

  • 規制の明確化 – PresenceがEU AI法やその他の新たな枠組みにどのように準拠するかに関するOpenAIからのアップデートは、より広範な導入に向けた試金石となるでしょう。
  • 価格モデル – プラットフォームおよびエンジニアリングサービスの料金に関する透明性が、このサービスが超大手企業以外でも実用的であるかどうかを決定づけることになります。
  • 拡大基準 – 「要件を満たす」顧客の定義や、サービスをより広い層に開放する計画は、AIインフラとしての領域における競争力学を形作ることになるでしょう。
  • 業界の反応 – AIモデルと現場のエンジニアリング人材を融合させたサービスに対し、従来のエンタープライズ・ソフトウェア・ベンダーやクラウドプロバイダーがどのような立ち位置をとるかが注目されます。

まとめ

OpenAIのPresenceは、AIエージェントと専任のエンジニアを組み合わせることで、ボットを本番環境で利用可能なコンポーネントへと進化させ、多くのボットを実験室の段階に留めてきた統合と信頼性の問題に対処します。このアプローチは大企業におけるAI主導の自動化を加速させる可能性がありますが、コスト、規制遵守、そして長期的なベンダー依存に関する疑問も投げかけています。今後数ヶ月間で、このモデルが一部の初期導入者を超えて拡大できるのか、あるいはニッチで手厚いサポートを必要とする限定的なサービスに留まるのかが明らかになるでしょう。