OpenAI Presence turns to consulting model, deploying AI agents through engineers instead of APIs

OpenAI Presence Shifts Focus to Enterprise Consulting

OpenAI has introduced OpenAI Presence, marking a strategic pivot from self-service APIs to a consulting-driven enterprise platform for deploying AI agents. Unlike previous releases, Presence is available exclusively through limited general availability, with deployments led by forward-deployed engineers from OpenAI and select global systems integrators rather than as a self-serve product 1

1

. This approach mirrors traditional consulting firms like Accenture and Deloitte, signaling OpenAI's intent to capture enterprise revenue through high-touch deployments 4

4

.

The platform focuses on real-time experiences across voice and chat, supporting AI customer support agents, outbound sales, and internal workflows. Companies can deploy AI agents to handle customer refunds, billing issues, account deletions, and order status queries while maintaining governance controls over agent behavior 1

1

. Each deployment addresses a single task, with agents receiving only the knowledge and system access needed for that specific function 3

3

.

How Organizations Build and Get Better Integrated with AI Agents

Presence provides an integrated control interface featuring an agent editor, agent playground, and simulation testing tools. Organizations can test agents against common requests, edge cases, and high-risk scenarios before production deployment . Companies determine what agents can do independently, which actions require human approval, and when escalation mechanisms should transfer interactions to human workers .

The platform addresses policy enforcement by allowing enterprises to set boundaries around agent behavior. Guardrails intervene when interactions move outside defined parameters, while graders evaluate whether agents reached intended outcomes, followed policies, used tools correctly, and escalated appropriately 3

3

. This human-in-the-loop approach gives organizations control over how realtime voice agents and chatbots represent their brand across customer interactions.

Codex Powers Continuous Improvement Loop

What distinguishes Presence from static chatbot solutions is its continuous improvement loop powered by Codex, OpenAI's coding model. After deployment, Codex investigates agent behavior, analyzes production sessions and escalations, then proposes code-level updates to improve performance . Teams test proposed changes against the production version before approving controlled rollouts, ensuring agents adapt to evolving company policies, products, and customer behavior without unchecked automated rewrites 3

3

.

OpenAI reports that Presence already powers its English-language phone support channel at 1-888-GPT-0090, resolving 75% of inbound issues without human assistance. The Codex-driven improvement process reportedly reduced human handoffs by 15 percentage points over just 10 days 3

3

. These figures, while company-reported and not independently validated, suggest meaningful operational gains from the automated improvement mechanism.

Early Adopters and Industry Reception

SoftBank, which has committed $60 billion to OpenAI, is exploring how Presence can enable trusted customer agents for Japanese-language interactions. Tadahisa Murakami, VP and head of SoftBank's data and digital transformation division, stated that frontline teams rated the agent's Japanese conversations highly for natural and accurate quality 1

1

. Other early customers include BBVA in Mexico experimenting with voice support in regular operations, and IAG testing customer support during high-traffic periods like bad weather .

However, industry skepticism remains. Gartner predicted that by 2027, half of organizations planning to shift customer service to AI will abandon those plans. "While AI offers significant potential to transform customer service, it is not a panacea," said Kathy Ross, senior director analyst for Gartner's customer service and support practice. "The human touch remains irreplaceable in many interactions, and organizations must balance technology with human empathy and understanding" 1

1

.

Pricing and Availability Remain Unclear

OpenAI has not disclosed pricing details, with deployments scoped individually based on each customer's use case and implementation needs. "Broader pricing details to come as availability expands," an OpenAI spokesperson stated 1

1

. The company also hasn't clarified whether Presence supports models from providers other than OpenAI, including increasingly popular Chinese open-weights alternatives like GLM-5.2 and Kimi K3 3

3

.

This consulting-based approach matters because it represents OpenAI's attempt to generate substantial enterprise revenue amid massive datacenter buildout commitments and the path toward profitability. The shift from cheap, accessible APIs to high-touch enterprise deployments reflects the company's need to serve customers who "might actually pay for its services" 1

1

. Whether this model proves sustainable, particularly in heavily regulated sectors like finance and insurance where trust and control are paramount, will determine if OpenAI's consulting gambit succeeds where self-service APIs fell short on revenue generation.