Data Moats and Increasing Returns in the AI Economy

Nowadays, consumers are flooding AI with their data.  Their queries, and information uploaded to support them, are giving tech companies terabytes of personal data.  Consumers are uploading photos, shopping lists, resumes, and text threads.  Companies can use this data to create customer profiles designed for targeted marketing.  How many customers of the next decade will be able to resist targeted advertising using years of their AI queries and social media posts as analysis fodder?

What Are Data Moats?

In the business world, data moats are systems used to keep data gleaned or generated by a company within the company for its use in research, product design, and marketing.  This data, which accumulates over time, provides the company with distinct advantages as it analyzes and refines these terabytes into usable nuggets of information.  Within their respective data moats, companies use their proprietary data to create ideas for new goods and services, marketing campaigns, and investment opportunities.

Data Moats Collecting User Data

Most large companies today do a majority of their business online, collecting data from consumers through their profiles, purchase histories, and website searches.  This data is valuable in gaining insights into consumer habits, desires, and liquid resources.  As a result of this data value, companies want to prevent competitors from utilizing it also.  Competitors who can access rivals’ data may use it to create desirable substitutes, siphoning customers from their rivals.  For example, a firm may use access to rivals’ data to find customer complaints about rivals’ products and intentionally create substitutes that mitigate these weaknesses.

Economics of Data Moats

By continuously collecting and analyzing consumer data, especially with the help of AI, firms can make significant increases in product desirability and price discrimination.  

Product Desirability and Demand Enhancement

Customer website searches, communications with staff, and reviews or complaints can be quickly analyzed with AI to determine which product or service features they most like or dislike.  AI can then further determine which customer-desired changes could be made most quickly and efficiently to boost sales.  To help maximize the benefits of the data moat, the firm would seek to patent or copyright these AI-identified improvements.  This would prevent rival firms, who have not yet identified these desired changes, from making them in the same manner.  Thus, firms that lag in the collection of customer data will lag more significantly in the ability to make meaningful, revenue-boosting changes to their goods and services.

Price Discrimination

While using AI to analyze customer data for the purpose of continuous product and service improvement is not very controversial, using AI to charge customers prices closer to the maximum they are willing to pay could be considered somewhat unethical.  Firms can use customer purchasing data, especially frequency of purchases after changes in price, to engage in price discrimination by charging certain customers higher prices based on AI analysis of individual customers’ willingness to pay.  Customers’ individual profiles can be used to analyze their willingness to pay, combined with demographic data.  AI can then search information on these customers to estimate their net worth (wealth) and income.

When customers purchase through their individual accounts, AI can use data analysis to instantly set the price to the highest each customer is likely willing to pay.  For example, it will automatically raise the price for a frequent purchaser who is analyzed as having both a high net worth and high annual income.  Conversely, to make a sale that is still profitable, AI might slightly lower the price for a new customer or occasional purchaser who is identified as having lower income.  This results in greater profit for the firm by reducing consumer surplus and transferring it to the seller.

Barriers to Entry

The increasing use of proprietary data sets and AI analysis of those data sets by large companies will create barriers to entry for smaller firms.  Efficiency gains and increased profit caused by price discrimination - both due to AI - will result in significant economies of scale for large, established firms that can reinvest their profits into new capital goods like factory equipment.  Start-ups will struggle mightily to compete with the advantages of these established firms, which have used AI to automate and “supercharge” research-and-development (R&D), production, marketing, and sales.

Start-ups that do not have data sets with which to work will have to begin operations with a “trial-and-error” approach to pricing, marketing, and product design.  While some start-ups may get lucky and begin with desirable goods and services, but many others will flounder.  Their initial output, not smoothly evolved by terabytes of analyzed customer data and reviews, will be seen as less desirable than substitutes from larger, more established firms.

Policy Implications for Competition and Innovation

Is it considered anticompetitive behavior for firms to use data moats?  Likely not, as firms have accumulated this data themselves.  However, if firms attempt to steal user data from other firms, such as contractors or third party firms through malware or digital back doors, that could be considered fraudulent.  Because anticompetitive behavior is illegal, legislatures and regulators will need to update regulations as AI offers new avenues for firms to unethically glean data from rivals, either through direct hacking or through indirect data mining.

This indirect data mining, also known as data scraping, can occur legally as a firm uses powerful software to scroll through a rival’s website.  Although the firm is not accessing a rival’s internal data, is its ability to comb through all public-accessible webpages of a rival company considered unethical?  After all, only large, established firms will likely have these capabilities, while start-ups will not due to their minimal resources.  It is possible that regulators will decide to ban the use of AI to perform data scraping, although this would be extremely difficult to enforce.  

One major challenge in limiting AI data scraping by firms is the fact that modern AI behaves like humans, giving large companies considerable ability to claim that their human employees, rather than AI, are visiting competitors’ websites.  Realistically, therefore, it is almost impossible to level the data analysis playing field between large firms, particularly oligopolists, and small start-ups.