OpenAI and PwC put AI agents in the CFO's office

Finance has always been the department most terrified of putting AI in charge. While a hallucinated marketing email is embarrassing, a botched tax filing or an overlooked compliance clause can trigger an audit nightmare.

Now, OpenAI is teaming up with accounting giant PwC to tackle the CFO's office head-on. The two companies announced a collaboration to build and deploy AI agents across enterprise finance rhythms, from forecasting and reporting to procurement, treasury, tax, and accounting close.

Eating their own dog food

Instead of just peddling theory, OpenAI is serving as its own "customer zero," testing agentic workflows inside its internal finance organization before selling them outside.

And they shared some surprisingly concrete receipts. By using Codex to build automation tools, OpenAI's finance team managed to process 5x more contracts without hiring a single new person.

During its recent fundraising, the company also deployed an internal tool called IR-GPT to handle and coordinate more than 200 investor interactions. Currently, PwC and OpenAI are actively building a dedicated procurement agent inside OpenAI's finance group to capture lessons for future enterprise deployments.

The accounting rulebook meets AI

Getting enterprise CFOs on board requires more than smart prompts. It requires heavy-duty compliance and governance.

OpenAI brings the model muscle and developer tooling. Finance teams use Codex to spin up custom dashboards, spend trackers, and exception-management setups, while Workspace Agents and Connectors let AI pull the right enterprise context across existing software stacks.

PwC steps in to provide the operational guardrails. The consulting firm handles finance transformation, internal controls, and enterprise-grade implementation to take agents from internal prototypes to production environments.

They are even tackling AI cost management. The partnership aims to give CFOs direct visibility into token consumption, AI usage trends, and projected spend, treating enterprise AI adoption just like any standard operating expenditure.

Why it matters

For enterprise finance teams, this marks a clear turning point from using AI for casual brainstorming to trusting it with core ledgers. If you run a corporate finance group drowning in invoice reconciliation or month-end close stress, this is worth watching closely. That said, getting existing ERP data clean enough for an agent to read safely is still going to be a heavy, expensive lift.

For OpenAI, partnering with PwC is a brilliant land grab. The real barrier to enterprise AI adoption has never been model intelligence; it is institutional trust and risk aversion. Bringing one of the Big Four into the fold provides instant air cover against cautious corporate boards, putting serious pressure on traditional enterprise software incumbents.

Still, we think this setup carries real liabilities. Corporate accounting demands near-zero tolerance for error. If an agent overlooks an invoicing anomaly or misinterprets a tax provision, the finger-pointing between the model maker, the implementation consultant, and internal management will get messy fast. Until accountability lines are tested in court, human oversight will have to remain uncomfortably tight.