OpenAI named Dali Rajic its chief revenue officer on August 13, replacing Denise Dresser roughly eight months into her tenure. Rajic, whose résumé spans Wiz, Zscaler and AppDynamics, will build what OpenAI calls a disciplined, metrics-led global revenue organization. Dresser stays through a transition. Separately, Axios reports co-founder Greg Brockman is taking a larger operational role, pushing to position OpenAI ahead of Anthropic in enterprise adoption as the company prepares for a potential IPO.
No public evidence suggests enterprise sales have stalled. OpenAI reports more than one billion weekly active users, more than two million business customers—double the prior year—and enterprise revenue exceeding 40% of the total as of April.
A Widening ROI Distribution Inside OpenAI's Own Data
Enterprise Signals data released August 12 shows acceleration at the extremes. The top decile of enterprise customers by usage now generates 8.3× as many output tokens per active user as the median firm, up from 2.6× in January. Weekly plugin usage runs at 21% in frontier accounts versus 9% elsewhere. Since February, weekly enterprise Codex users surged 108× in legal, 41× in sales and recruiting, and 26× in marketing.
A narrow cohort has wired models into tools, permissions, company data and repeatable processes; their consumption is exploding. The median organization has distributed seats but done little else.
Deloitte's 2026 survey confirms the pattern: 48% of companies deployed AI without redesigning surrounding workflows, only 12% redesigned at scale, and 4% report AI value to boards at the highest maturity level. McKinsey's July work: only 21% of firms have rearchitected operating models around AI.
OpenAI's Multibillion-Dollar FDE Bet
In February OpenAI formed Frontier Alliances with McKinsey, BCG, Accenture and Capgemini. In May it launched its majority-controlled Deployment Company backed by more than $4 billion and acquired Tomoro, adding roughly 150 forward-deployed engineers. In June it committed another $150 million to a partner network of integrators and consultants.
If token access alone generated ROI, a model vendor would need little more than API endpoints and a billing system. OpenAI is building something closer to Palantir's forward-deployment playbook atop a consulting ecosystem: data integration, permissions, workflow redesign, change management, monitoring, agent deployment, outcome measurement.
The Competitive Unit Is Shifting
Google released Gemini 3.7 Flash on August 13 at $0.75/$3.75 per million input/output tokens—introductory pricing through December 31, 2026, after which rates double. OpenAI already offers a tiered ladder: GPT-5.6 Luna at $0.20/$1.20, Terra at $2/$12, Sol at $5/$30. Intelligence is becoming price-segmented and routable, and the unit under competitive pressure is migrating from cost per token to cost per successful completed workflow—inference plus retrieval, tool calls, retries, human review, monitoring and integration amortization.
Today's tape confirms differentiation over blanket de-rating: IGV up 2.94%, Salesforce +3.30%, Palantir +4.57%, ServiceNow +1.45%. Salesforce's Agentforce has crossed roughly $1.2 billion ARR; ServiceNow's AI products exceeded $1 billion annual contract value with agentic deployments growing ninefold in nine months.
Cheap Intelligence, Expensive Implementation
The 2026 evidence points toward a structural split: declining unit cost of intelligence paired with rising expenditure on data integration, governance, workflow redesign and orchestration. Seat adoption does not equal workflow adoption, and workflow adoption does not equal measurable ROI.
Model inference margins face compression while the orchestration layer—ServiceNow's workflow governance, Salesforce's CRM context, Palantir's data ontology, specialist FDE firms—can sustain premium pricing because those companies own the customer's actual working process. Informed capital is following accordingly: Cisco acquired Astrix for agent identity security, Cyera is paying a reported $1 billion for Oasis, Snowflake invested in AtScale's semantic-layer architecture. Capital is clustering around deployment, identity and workflow governance.
OpenAI's CRO swap reads as acknowledgment that enterprise AI requires forward-deployed engineers, partner ecosystems and organizational change capacity to monetize. The commercial architecture under construction—$4 billion deployment vehicle, 150 acquired FDEs, $150 million partner fund, Brockman's operational focus—amounts to a bet that embedding workflows today will ignite recurring agent consumption tomorrow. If that flywheel turns, the deployment layer becomes a moat. If it does not, OpenAI has converted software margins into consulting economics. That is the wager Rajic inherits—and the clearest signal yet of where enterprise AI profit pools are forming.
not investment advice