Baker Hughes Achieves Bottom Line Savings with AI Sourcing Optimization

Introduction

Baker Hughes is a global energy technology leader with 57,000 employees operating across 120+ countries. Generating $7B+ in quarterly revenue, the company provides innovative products and systems essential to the energy sector.

Despite this global reach, Baker Hughes recognized that their sourcing teams were hindered by a challenge common to industrial giants: fragmented data. Critical information was scattered across disconnected internal and external systems, forcing buyers to manually reconcile data through multiple spreadsheets. This “one-off” approach slowed procurement velocity and limited the team’s ability to manage their portfolio with speed and confidence.

In 2019, Baker Hughes partnered with C3 AI to embed artificial intelligence into the core of their supply chain. The goal was to unify scattered data, surface invisible insights, and enable higher-quality decisions earlier in the procurement process. What began as a pilot across three sites quickly scaled to ten, and finally across the enterprise. This collaboration evolved beyond a simple dashboard into a comprehensive sourcing decision system – an intelligent layer that allows teams to negotiate with clarity in a volatile environment.

Challenges

Procurement and sourcing managers operate in an increasingly complex environment where global supply chains, cost pressures, and risk management intersect. Beyond negotiating favorable terms, they must ensure continuity, compliance, and resilience across diverse supplier networks. While the sourcing teams have deep industry experience, they were frequently “buried” under fragmented data. The manual nature of their workflow created several critical pain points:

Lagging Insights: By the time data points were analyzed and actioned, purchase orders exceeded the cancellation period or market movements rendered the analysis obsolete.

Systemic Discrepancies: Different ERP systems occasionally shared identical part numbers for different items, while foreign exchange (FX) fluctuations created “phantom savings” that did not exist.

Cultural Hurdles: Ground teams sent purchase orders for quotes as mere placeholders to meet order cycle fulfillment targets, further obscuring the true data landscape.

Acknowledging these gaps early was pivotal to addressing underlying issues for a solid foundation to implement AI-driven approaches.

The Solution

Through the partnership with C3 AI, Baker Hughes successfully transformed fragmented manual processes into an automated, high-confidence decision system. Opportunities are presented with a full evidence package – total landed cost breakdowns, similar part rationale, and index-based pricing signals. The user feedback loop created a system that enables users to train future recommendations with “internal knowledge” by accepting or rejecting opportunities and connecting live market indices to SKUs. Buyers can monitor markets in real time through live tariff feeds to immediately estimate the total landed cost of an order as new executive orders are announced.

High-Volume Capitalization: By focusing on high-volume product lines and facilities, the team achieved a 5–7% acceptance rate on identified opportunities which significantly impacted the cost of goods sold, which is typically in the high seven- to low eight-figure range annually.

Weekly Rhythms: Success was sustained through weekly business reviews where teams shared live feedback and captured savings in real time.

About Baker Hughes

• $7+ billion in quarterly revenue
• 120+ countries of operation
• 800+ program facilities worldwide
• 57,000+ employees

Project Highlights

• 30+ million records from 5 enterprise data sources integrated
• 200+ active users
• Realized savings validated and accepted by business users
• Mobile enabled (iOS & Android)
• Comprehensive view of total landed cost including logistics and live tariff rates
• Identification of similar parts for leverage in supplier negotiations
• Human-in-the-loop workflow to accept or reject recommendations
• Improved recommendations based on continuous user actions