Dynamic pricing algorithms offer competitive advantages but come with mounting regulatory risks. Adam Berry, Mike Horoho, Andrea Levine and Alex Tully of FTI Consulting outline how these technologies have transformed industries while attracting scrutiny from government enforcers and civil plaintiffs.
Adam Berry, Mike Horoho, Andrea Levine and Alex Tully co-authored this article.
In recent years, the use of AI and algorithms in business operations has surged, transforming industries from consumer finance to retail goods. Among the most significant developments is the widespread adoption of pricing algorithms, which enable companies to adjust prices dynamically based on market conditions, consumer behavior and competitor strategies.
While these technologies promote efficiency and competitive advantages, they also raise critical concerns. Data privacy risks emerge as algorithms rely on vast amounts of consumer information, often with limited transparency. Consumer protection issues arise when pricing strategies lead to discrimination or exploit behavioral biases. Equally concerning is the potential for algorithmic collusion — where independent companies’ pricing algorithms, without explicit agreement, leverage non-public competitively sensitive information or otherwise learn to coordinate prices in ways that harm competition. As government enforcers and civil plaintiffs raise these challenges in the court system, the need for greater oversight and accountability in algorithmic decision-making is becoming an increasingly important risk management priority for businesses.
AI and algorithms
AI-related technology has unlocked tremendous opportunities for innovation, and its impact has been well-documented. Today, terms like AI, machine learning, natural language processing and deep learning are commonplace, discussed by mathematicians and school children alike. As AI capabilities become central to corporate strategy, decisions regarding algorithmic decision-making have taken precedence across industries.
It’s easy to understand why. AI-driven capabilities — such as demand forecasting, gaining deeper insights into price elasticity and applying dynamic pricing — have the potential to fundamentally reshape how a company operates. However, innovation is not void of risk. It challenges established norms and explores uncharted territory in the pursuit of discovering new and improved methods or solutions. As companies continue to adopt AI-driven strategies, it’s imperative that they understand the associated risks — the blind spots and subtle biases — including anticompetitive behavior.
Framework to addressing algorithmic risk
To comprehensively understand risks rooted in an algorithmic decision-making process, companies should adopt an interdisciplinary approach — one that integrates business, technical and legal perspectives. This holistic approach ensures that business impacts, legal risks and technical underpinnings are understood and pave the way for both informed decision making during initial deployment and continuous monitoring once these tools are in production. It also provides organizations with a structured framework for remediation in instances where an algorithmic decision-making process is challenged by regulators, civil plaintiffs or internal stakeholders.
At a minimum, an effective algorithmic risk management framework includes:
- Stakeholder interviews: Conversations with business leaders, compliance officers and data scientists offer qualitative insights, the chance to validate technical findings and practical perspectives on how each algorithm operates in the real world.
- Technical documentation: Careful examination provides critical context about system design, underlying assumptions, intended functionality and limitations of the algorithm.
- Code review: This involves using software tools that facilitate analysis of the algorithm’s structure and logic to help uncover potential flaws, biases or inefficiencies in the code.
- Data lineage assessment: Tracing the flow of data from its source to the point of decision-making ensures transparency in how data transforms through various systems.
- Exploratory data analysis: Identifies issues with data quality and informs teams of the range of values for each input. This should be done in a dedicated analytics environment.
- Statistical techniques: Helps evaluate the fairness, accuracy and performance of the decision-making process. This is vital for analyzing the outputs and with the above tools and creates a comprehensive approach that ensures each step contributes to a robust understanding of the algorithm’s impact.
Applying the framework to algorithmic anticompetitive risks
As businesses begin to more actively integrate AI into their business operations, enforcement agencies and private plaintiffs have begun to scrutinize the impact of these technologies on competition. When evaluating the antitrust risks associated with algorithmic decision making and other advanced technologies, this framework helps determine whether the software reflects legitimate, unilateral business decisions or, as argued by civil plaintiffs and the DOJ in certain industries, serves as a proxy for coordinated activity among competitors.
These arguments and related court decisions have focused on the combining of sensitive non-public pricing and supply information, the extent to which companies delegate decision-making authority to the algorithm and the impact of the use of the algorithm on prices to consumers. A particular focus has been algorithms embedded in third-party pricing software.
Proper antitrust risk mitigation would, therefore, involve an understanding of the data inputs and outputs to the pricing optimization software, the extent to which the software has access to competitively sensitive information and whether that information plays a role in the pricing recommendations made.
Companies should also be mindful of the extent to which they maintain decision-making authority including the use of a standard algorithm versus making company-specific adjustments to achieve unique business objectives and the ability for employees to override pricing recommendations.
Lastly, in evaluating the performance of the software, companies should be mindful of any trends in pricing, margin or supply that do not align with current market dynamics. Such trends may be indicative of unintended alignment with other software users.
Conclusion
We are at a fundamental inflection point where the opportunities for innovation offered by AI are immense. But so are the risks. A comprehensive compliance framework allows companies to obtain a complete, fact-based understanding of the risks associated with its algorithmic decision-making processes, including that of algorithmic anticompetitive behavior. It enables companies to embrace and invest in algorithmic decision-making capabilities knowing the associated risks are being properly managed.
