OpenAI and PwC Team Up to Rebuild the Office of the CFO Around AI Agents

OpenAI and PwC are collaborating to rebuild the office of the CFO around AI agents, with OpenAI's own finance organization acting as "customer zero" for workflows the partners intend to bring to enterprises at scale.

The collaboration, announced by the two companies, targets the core operating rhythms of finance: planning, forecasting, and reporting on one end, and procurement, payments, treasury, tax, and the accounting close on the other. The partners say agents can automate repeatable work, coordinate across systems, surface risks and insights, and support decisions — all under governance with human oversight.

The stakes for the enterprise market are straightforward. Finance teams sit at the center of how organizations plan, allocate capital, manage risk, and make decisions, according to the announcement. If agentic AI can reliably handle contract review, payment monitoring, and forecast updates, it changes the cost structure and speed of one of the most conservative, control-heavy functions in the enterprise. That is why a Big Four firm pairing with a frontier model provider matters beyond the press release: it signals that agent workflows are moving from demos into production environments where controls and auditability are non-negotiable.

Building agents around real finance workflows

The partners emphasize what sets the effort apart: building in the real world rather than designing in theory. The most concrete example is a procurement agent that PwC and OpenAI are building inside the OpenAI finance organization. Learnings from that build will feed additional agents across core finance workflows.

OpenAI has already deployed AI agents across its own finance function, using ChatGPT and Codex to support work in investor relations, treasury, tax, reporting, corporate development, and contract review. The new collaboration extends those internal learnings to other CFOs.

In practice, the partners describe a set of specific agent tasks:

• Monitoring payments and exceptions
• Reviewing contracts or invoices against policy
• Updating forecasts as business conditions change
• Preparing reporting materials
• Surfacing risks before month-end or quarter-end close

The product stack matters here. Codex can help teams build and adapt the tools behind these workflows — dashboards, spend trackers, and exception-management systems. Workspace Agents, recently introduced in ChatGPT, can make those workflows repeatable across the tools teams already use. Skills and Connectors let agents follow approved processes and work with the right enterprise context.

PwC's role is the transformation layer: finance expertise, controls, and implementation muscle to move workflows from prototype to production in real enterprise environments.

OpenAI as customer zero

The OpenAI finance organization is serving as "customer zero" for applying AI agents to finance work, using that production environment to validate enterprise-scale workflows, governance models, runtime controls, and human-agent collaboration patterns.

The internal results the company reports are concrete. The finance team used Codex to process five times more contracts with the same-sized team. It also used an internal tool called IR-GPT to manage more than 200 investor interactions during the recent fundraise.

Those numbers carry weight because they come from a production finance function, not a benchmark. A 5x throughput gain in contract processing with unchanged headcount is the kind of figure CFOs will ask their own teams to replicate — and the kind that will drive scrutiny of how the workflows were governed.

The governance angle

The announcement spends significant space on controls, which reflects where enterprise AI deployments actually stall. As agentic workflows scale, CFOs will need visibility into AI usage, token consumption, and projected spend, the partners note, so finance teams can govern adoption the same way they manage other enterprise operating costs.

That framing positions AI spend as an operating expense line to be managed, not a science project. It also previews a likely next battleground for enterprise AI: usage observability. Finance leaders who cannot see what agents are doing — and what they cost in tokens — will struggle to approve wider rollouts.

PwC frames the shift as structural rather than incremental.

"Finance is at an inflection point, where organizations are moving from process efficiency to intelligent, decision-centric operations. Through our collaboration with OpenAI, we're helping clients embed agentic AI into the core fabric of the finance function, enabling more proactive insights, stronger controls, and a more adaptive operating model."

— Tyson Cornell, US Advisory Leader, PwC

Division of labor

The partnership's structure follows a now-familiar pattern in enterprise AI: the model provider supplies capability, the consultancy supplies deployment. OpenAI brings the models and product capabilities that make it possible to build, adapt, and govern agent workflows across the platforms finance teams already use. PwC brings finance transformation, controls, and implementation expertise.

Together, the partners say, they are turning internal learnings from AI-native finance into a practical path for CFOs to deploy agentic workflows across their organizations.

The "customer zero" playbook is itself part of the story. OpenAI validating workflows, governance models, runtime controls, and human-agent collaboration patterns inside its own finance function gives PwC reference material that predates any client engagement — and gives prospective buyers a live example rather than a roadmap slide.

What to watch

The announcement leaves open questions the market will answer over the coming quarters: whether the procurement agent built inside OpenAI transfers cleanly to other enterprises, whether token-consumption governance tools mature fast enough to satisfy audit requirements, and whether CFOs trust agents enough to let them touch the accounting close.

The direction of travel, though, is clear from the source material. The collaboration moves agentic AI from efficiency pilots toward decision-centric finance operations, and both partners are betting that the office of the CFO — long a laggard in automation because of its control requirements — becomes one of the first enterprise functions rebuilt around agents rather than bots executing scripts.