pwc
PwC × OpenAI Collaboration

Enabling the AI-Native Finance Organization

PwC and OpenAI are collaborating to reimagine the office of the CFO — building AI agents around the core operating rhythms of finance, from planning and forecasting to procurement, payments, treasury, tax, and the accounting close.

"Finance is at an inflection point, where organizations are moving from process efficiency to intelligent, decision-centric operations."

— Tyson Cornell, US Advisory Leader, PwC

01

AI for Hybrid Workforce

Custom + Oracle Agents creating Hybrid (Agent + Human) workflows

02

AI as Front Door

ChatGPT gateway for all enterprise applications and orchestration

03

AI-Driven Managed Services

PwC AI Native Managed Services for Finance & HR on OpenAI on OpenAI

1

AI for Hybrid Workforce

As part of OpenAI's Oracle implementation, PwC is leveraging Custom + Oracle Agents to create hybrid (Agent + Human) workflows across the finance function.

Oracle ERP & EPM

Digital core systems

Custom + Oracle Agents

AI-powered automation

Human Review

Oversight & approval

Orchestration

End-to-end coordination

25
AI-Powered Steps
11
Human Steps
11
Oracle Embedded
14
APP Agents

Record to Report

Master Data Maintainer — 65% cleaner master data
Smart JE Preparer — 70% reduction in manual JE prep
GL Close Accelerator — 45% faster anomaly detection
Flux Analysis & Narrative Generator — 55% faster prep

General Ledger Close

System Closer — 75% faster close cycle
Auto Reconciliation Agent — 80% auto-matched reconciliations
Close Task Orchestrator — 40% fewer missed deadlines

Consolidation & Reporting

TB Virtual Assistant — 60% faster internal reporting
Consolidation Optimizer — 50% faster adjustments
Audit PBC Pack Compiler — 60% less time gathering evidence
CFO Digital Assistant — 45% faster CFO reporting prep
2

AI as the Front Door

PwC is helping OpenAI implement a ChatGPT gateway for all enterprise applications. All user interaction, orchestration, and insights will be exposed through the LLM Gateway — powered by MCP.

OpenAI Northstar Vision — Agentic Bets for 2026

ChatGPT/Codex as the Front Door

One interface to access context, insights, and actions across enterprise systems.

Autonomous Agents at Work

Agents execute background workflows automatically or on demand, with HITL controls.

Bespoke Apps with Codex

Rapidly build internal tools where speed and fit matter more than buying software.

Connecting Platforms to OpenAI via MCP

HRIS
GTM
Planning
Internally Built Platforms
ERP
Data Lake
MCP
Questions
Insights
Actions
Visualizations
ChatGPT
Codex
Workspace Agents

Finance Context Protocol

Reference architecture for a first-of-its-kind AI Native Finance Function — the governed flow of context from request to action.

Finance Context Protocol
The governed flow of context from request to action
1
AI Finance Agent
Interact & Orchestrate
AI Finance Agent
Understand Request
Gather Context & Find Information
Coordinate Tasks & Orchestrate Workflow
Communicate & Engage
Provide Explanations & Recommendations
2
Evidence & Facts Layer
What is true
Evidence Sources
Invoices
POs
Receipts
Contracts
Emails
Policies
Fact Creation & Enrichment
Extract
Normalize
Link Entities
Validate
Trusted Finance Facts
SupplierInvoicePOReceiptContractTaxPayment TermsBankApprovalRisk
3
Deterministic Business Decision Layer
What is allowed
Finance Rules & Controls
  • • Approval Limits (DOA)
  • • Tolerances
  • • Segregation of Duties
  • • Tax & Compliance Rules
  • • Payment Eligibility Rules
Risk & Value Scoring
  • • Fraud / Duplicate Risk
  • • Supplier Risk
  • • Payment Risk
  • • Early Pay Discount Value
  • • Exception Likelihood
Optimization Engine
  • • Best Allowed Action
  • • Payment Timing
  • • Working Capital Impact
  • • SLA & Capacity Constraints
  • • Cost / Risk Trade-offs
Recommended Action
  • • The best action allowed by rules and constraints
4
Decision Ledger
Why it happened
Facts Used
What data and evidence informed the decision
Rules Applied
Which rules passed, failed or were triggered
Decision Made
Recommended action and rationale
Human Feedback
Approvals, overrides, comments and justifications
Actions Taken
What was executed, by whom and when
Audit Trail
Immutable history for compliance and audit
5
Systems of Record & Execution
Make it real
Enterprise Systems (Systems of Record)
ERP
AP System
Procurement
Supplier Master
Payments / Treasury
Tax System
JE / Subledger / GL
Workflow / Case Mgmt.
Execution Examples
Place invoice on hold
Route for approval
Approve & release payment
Create journal entry
Update supplier record
Create exception
Security & Access
Role-based access, data encryption, and least privilege across all layers.
Governance & Change Management
Rules, reference data and models are governed, versioned and reviewed.
Monitoring & Continuous Improvement
Monitor outcomes, measure performance, and use feedback to refine rules and models.
Faster Cycle Times

Automates coordination to accelerate decisions

Stronger Controls

Deterministic rules and audit trails

Better Outcomes

Optimized payment timing & working capital

Full Transparency

End-to-end decision visibility

Empowered People

Focus on exceptions & value-add

3

AI-Driven Managed Services

For OpenAI's DeployCo, PwC is building AI Native Managed Services for Finance and HR — powered by OpenAI on OpenAI — transforming service delivery from reactive to proactive.

Smarter Service Delivery

  • Always-on monitoring with proactive alerts
  • Playbook-driven resolution for consistent outcomes
  • Continuous learning to reduce manual effort

Connected Control Tower

  • Real-time ERP signal ingestion for data-driven decisions
  • Scalable automation with human-in-the-loop precision
  • Cross-tower insights connecting GL, AP, and AR

Supercharged Business Outcomes

  • Faster close with fewer anomalies, quicker reconciliations
  • Working capital gains — faster invoicing, reduced DSO
  • Spend control for stronger compliance, less leakage
  • Better commercials — transparent KPIs, fewer tickets
Improved KPIsClose cycle timeContract coverageInvoice resolutionRevenue accuracy

AI-Driven Managed Services — Structural Reimagining for Growth

Full structural redesign — AI as the orchestrator, operating model and role redefinition to outcome-oriented jobs. Humans govern outcomes, not execute tasks.

Target: PE-backed companies with value creation mandates, or cost-optimization cultures

Organizing Construct
Outcomes → Triggers → Actions
Design Principle
Reimagine from AI-first principles
Operating Model
New model — AI-native from day one
Finance Role
Humans govern; AI executes & manages
Value Case
Structural cost takeout + ecosystem Finance

AI-Native Finance Roles

Policy Manager

Owns policy models, rules, controls. Ensures policy content is approved and language is accurate.

Decisions & Outcomes Manager

Owns performance, decision quality and economic outcomes. Reviews how conversations lead to decisions.

Agent Workforce Manager

Owns agents, tools, permissions and operating performance. Monitors agent behavior, routing, handoffs.

Activation Sequence — AI-Driven Managed Services

Model, Blueprint, Scale — the journey to an AI-native Intelligent Enterprise.

0Learn

Finance Strategy

Understand the landscape, define the ambition, build conviction

1Model

Finance Intelligence

Digital twin modeling — current state, cost-optimized, and AI-native scenarios

2Blueprint

Finance Scaffolding

Design the agentic organization with governed blueprint

4Scale

Intelligent Enterprise

Full structural redesign — AI-managed, human-governed

Governed by Finance

Policies and controls are owned by Finance, not the AI.

Continuously Improved

Insights and feedback drive ongoing improvements.

Auditable and Compliant

Every decision is traceable, explainable and accountable.

pwc

© 2025 PwC. All rights reserved. PwC refers to the PwC network and/or one or more of its member firms, each of which is a separate legal entity.

Confidential — For client discussion purposes only.

pwc.com