An AI solution for anti-corruption compliance at the world’s largest brewer

Summary:

The AI solution has reduced the company’s costs associated with the broad topic of anti-corruption compliance by millions of dollars. 

Client:

AB InBev is the world’s largest brewer, operating as a holding company involved in the production and distribution of both alcoholic and non-alcoholic beverages. Founded on August 2, 1977, the company is headquartered in Leuven, Belgium.

Problem Statement:

In 2016, AB-InBev made a $100 billion acquisition of SABMiller. While finalizing this deal, AB-InBev was also resolving an FCPA investigation related to bribery offenses in India.

With two simultaneous objectives—settling the FCPA probe and preparing for post-merger integration—AB-InBev settled the FCPA investigation two weeks before the merger was completed. As part of the settlement, AB-InBev agreed to pay $6 million for violating the Foreign Corrupt Practices Act (FCPA).

Additionally, the company committed to cooperating with the SEC and reporting its FCPA compliance efforts for two years. This settlement added ongoing reporting obligations to AB-InBev just as it was acquiring SABMiller, a complex enterprise with numerous decentralized operating units, many of which operated in high-risk jurisdictions. 

Traditionally, companies assess the compliance practices of acquired businesses through extensive interviews with local employees conducted by lawyers or auditors flown around the world. AB-InBev executives, however, found this approach inefficient and sought a more data-driven method. By consolidating all relevant data, the company aimed to make more informed management decisions and enhance compliance practices.

Results: 

• A 86% reduction of the company’s costs associated with investigating suspect payments from about $1.8 million to about $250,000.
• Successful identification and recovery of funds that could have been lost due to fraud, corruption, or inadequate oversight. 
• In some countries, the ML model can predict with an approximate 80 percent confidence level whether a vendor has a connection to a government official.

AI Solution Overview:

AB-InBev has developed ML technology to identify risky business partners and potential illegal payments. The BrewRIGHT analytics platform was created to consolidate data from finance, compliance, human resources, and other systems, enhancing the detection of transactions and third parties posing risks. Drawing data from operations in over 50 countries, BrewRIGHT enables proactive monitoring of legal risks and prevention of violations.

The platform is built on a risk-scoring approach, where transactions or relationships are categorized as higher-risk based on various risk attributes (such as urgency of payment, payment to a political or state-owned entity, and vendor type) and the weighting of these attributes. Initially developed with external assistance, AB-InBev transitioned to an internally managed solution within a year, expanding the tool to cover operations across the entire enterprise.

Combining ML technology, including supervised learning, with the BrewRIGHT analytics platform, AB-InBev now provides users with dashboards built on third-party data-visualization software. BrewRIGHT addresses various compliance concerns including anti-money laundering, antitrust, conflicts of interest, payments, third-party vendors, and travel and entertainment. Importantly, all workflows generate structured datasets used to update and inform ML models.

References: 

1. AB InBev Taps Machine Learning to Root Out Corruption. https://www.wsj.com/articles/ab-inbev-taps-machine-learning-to-root-out-corruption-11579257001
2.  Using Machine Learning For Anti-Corruption Risk And Compliance. https://www.coalitionforintegrity.org/wp-content/uploads/2021/04/Using-Machine-Learning-for-Anti-Corruption-Risk-and-Compliance.pdf 
3. Talking Compliance Analytics at AB-InBev. https://www.radicalcompliance.com/2019/01/10/podcast-compliance-analytics-ab-inbev/

Industry: Anti-Corruption Compliance

Client: AB InBev

Publication Date: 2016