Corporate bribe payers should think twice in the era of artificial intelligence and big data

…and it starts with Brazil

Anti-corruption enforcement has been intensifying across Latin America ever since the Operation Car Wash investigations made headline news in March 2014. But Operation Car Wash is just one of several major regulatory probes involving state-owned entities, private companies, politicians, lobbyists and other intermediaries accused of wrongdoing in a number of industries, including oil and gas, electricity, financial services and infrastructure. And while the Operation Car Wash investigations may be winding down in Brazil, their effects are increasingly being felt throughout Latin America. In countries such as Colombia, Peru and Argentina, other probes are just beginning to ramp up. 

Looking beyond the spectacular media reporting of the well-orchestrated police operations across Brazil and the steady year-on-year deterioration of the corruption perception index as reported by Transparency International, we see a silent revolution at a growing number of private and public sector companies. Compliance is now a board-level issue for many major Brazilian companies. Implementation of effective programs to prevent, detect, and remediate fraud and corruption risks is now a top priority for boards and executive teams alike.
Technology is set to radically alter the accuracy, speed, and efficiency of future compliance programs.

 The arrival of big data and artificial intelligence technology in the compliance, anti-corruption, and anti-fraud space is a real game changer. Top Silicon Valley companies and financial institutions have used these technologies for many years, but a strong business case for their use had evaded the compliance function. Brazil, however, is now a fertile ground for the compliance-related adoption of these technologies. There are several important reasons for this. First, most companies in the country are starting from scratch with their programs, which means they can adopt the latest technologies without disrupting incumbent processes and systems — sort of like jumping from faxing documents to 5G data transmissions. Second, Brazil is a tech-savvy culture with some of the highest internet penetration levels and social networking density in the world, creating a natural environment for new technologies to prosper. Third, due to having one of the most cumbersome bureaucracies in the world, Brazilians have become creative at using technology to overcome such obstacles. The fact that these technological solutions have become more powerful and cost-effective makes their adoption much more palatable for companies that have been in cost-cutting mode while navigating the longest recession in Brazilian history. Finally, anti-corruption watchdogs in Brazil are in learning mode and increasingly receptive to technology- and analytics-enhanced compliance solutions. 

So how can technology help corporations deploy a digital dragnet to catch bad actors? Almost everything has a digital footprint in today’s world: bribes involve money, and money leaves a trail. Just as everything lives forever on the internet, a corporation’s financial transactions and employees’ communications, decisions, and behaviors are captured in databases, servers, phones, the cloud—the list goes on. The idea is to essentially hoover up broad sets of data available to the company, including financial data, emails, chat applications, HR databases, social networks, and public records/media. With this data, companies can run analytics to identify and correlate potentially problematic areas or issues, enabling some level of predictive capability. These data models and algorithms can be customized by corporations to flag any unusual patterns, trends, and outliers for human review, getting ahead of potential issues before they manifest into serious problems.

Artificial intelligence can automatically learn from previous instances of bribery and fraud and look for data with similar characteristics. It’s no longer just about finding the needle in the haystack, but automating the ease and efficiency of finding similar needles. For example, a group of executives implicated in one corruption scheme may have orchestrated other as-yet-undiscovered schemes, so it benefits the corporation to check for similarly suspicious patterns to ensure a clean house. Compliance officers can leverage institutional knowledge of existing control weaknesses and surmise how someone with an intent to defraud would take advantage of those gaps. In this scenario, an investigator could define the evidentiary data profile of fraudulent transactions and ask the AI program to query the data for matching records. This type of “machine learning” grows more powerful the more it is used and the more data that is fed into it. Each iteration generates feedback that improves accuracy and precision.

Advances in data analytics and increasingly interconnected people, tools, and services are allowing investigators to triage and analyze a large amount of previously unavailable and disparate information. For example, using data from accounting systems, email servers, and public business databases, an analyst can map communications to third-party individuals who are beneficiary owners of shell companies that have received suspicious payments or funneled bribes to an official. Location data from cell phones can identify whether individuals were near alleged meeting locations of cartel members engaging in anti-competitive behavior. Monitoring devices on corporate computers can detect when employees access suspicious websites and record keystrokes or take screenshots. Admittedly, these technological advancements do carry implications for personal privacy. However, when used appropriately and legally, with properly configured anonymization settings, they can be powerful deterrents against fraud and corruption.

As data volumes grow exponentially and geographic footprints expand, compliance departments need more than just additional headcount to effectively monitor problematic behavior. The deluge of information will overwhelm what a single individual, or even a team of individuals, can process. A technology-enabled solution will deliver greater productivity, accuracy, objectivity, and speed of analysis; prioritize mitigation and remediation measures; and allow skilled and finite personnel resources to focus on higher-level tasks. 

We are several years away from full-scale artificial intelligence—human judgment, ethics, and intervention will still be required in the fight against fraud and corruption for at least the near future. But the level of sophistication has changed, and the business case is solid. The next time an executive pledges to take on corruption, machine learning will be a powerful tool to help fulfill those promises.

• This article first appeared in Ethisphere’s Special Report on Latin America.