• ALEX ELIAS
• 12:12 PM ET

"Monica Kusaka in the redbrick brownstone on N. Damen Ave. across from the park, this lamp is for you. It combines your two favorite things: warm lighting and llamas." This was a real advertisement from retail brand Cost Plus World Market. Such overly personalized ads alienate many shoppers. Three in four consumers say highly personalized ads, as well as marketers' requests for personal information, are "creepy.

"Despite the risk of spooking customers, most marketing departments still closely monitor customers' purchase and browsing histories. For more than a decade, this has been the only way to construct accurate customer profiles and send them relevant ads. Fortunately, there's now a better way.

Thanks to artificial intelligence, marketers can create targeted ads based on just a few anonymous data points. This "probabilistic" marketing model generates strong sales without creeping out customers or exposing firms to legal and reputational risk in the event of data breaches. For years, marketing departments have been caught in a Catch-22.

If they didn't collect enough data to personalize customers' shopping experiences, they'd forego billions in revenue. In 2016, businesses collectively missed out on $756 billion in sales by failing to personalize, according to Accenture.But if they did collect customer data, firms risked losing customers' trust.

Consider what happened to Facebook (FB) when 87 million users learned that Cambridge Analytica, a political consulting firm, secretly scraped their data from the social media site.After the story broke, the company lost $50 billion in market value in just a few days. CEO Mark Zuckerberg was dragged to Capitol Hill to testify before Congress.

Brands can escape this no-win situation by changing how they target customers. Right now, most companies use "deterministic" data modeling. When a customer logs into a site with his username and password, a brand can follow his actions and suggest future purchases. With probabilistic data modeling, brands don't digitally stalk the customer.

Instead, they create an anonymous profile of a person with a few known preferences.For example, they may know that a certain customer is a 26-year-old male who enjoys science fiction. Based on that information, the brand can use artificial intelligence to infer what that anonymous customer will probably also like.

AI can be used to analyze millions of podcasts, books, hotels, designers, artists, and cafes and generate billions of cultural correlations. For instance, Spotify could use AI to learn that anyone who listens to a particular artist is also likely to shop at H&M. Spotify could then sell H&M the right to run ads whenever people play songs by that artist.

Probabilistic data keeps consumers' identities secure and also allows brands to gain insights from a much wider array of consumer-related topics and interests.Chipotle could find out which podcasts its users probably enjoy.

American Express could predict where a subset of customers will travel this summer on vacation. Nike could predict which Hulu shows its customers probably watch.

Probabilistic cultural correlations are highly reliable.Consider Amazon's "you may also like" recommendation engine. By leveraging probabilistic, anonymous data, Amazon can expose consumers who have clicked on certain products to similar products that might interest them. The engine drives over 30% of the tech giant's sales.

Or consider my company Qloo. Given specific preferences in one cultural domain, say food or music, Qloo's cultural AI can accurately infer relevance for a given entity in a different domain 55% of the time, on average. With two known preferences, that number jumps to 75%, and with three, more than 90% Soon, using anonymized probabilistic data may become a financial imperative.

The European Union's General Data Protection Regulations, which went into effect last month, aim to "give citizens back control of their personal data" by implementing a higher standard for data protection and privacy.

In theory, the rules only apply within the EU. But in practice, all multinational companies with a significant European presence must adhere to them, since it'd be difficult to separate the data of EU users from non-EU residents. The EU plans to fine companies up to 4% of their global annual revenue if they refuse to comply.

Luckily, aside from probabilistic data, companies have another modern data privacy tool at their disposal — blockchain. Blockchain decentralizes databases ranging from title insurance to social identities, so that no single corporate entity owns the data.

Essentially, it gives consumers private keys to their own data in a public domain. It won't be long before consumers will have complete control over which companies see data on their tastes and preferences and when, perhaps even rendering GDPR regulations moot.

Consumers don't like it when brands stalk them and pinpoint who they are, even if it eventually leads to a cool llama lamp. Eventually, blockchain will give consumers full sovereignty over their data. Until and even after then, companies can use AI to better divine consumers' preferences without ever collecting terabytes of data on precise identities.

• Elias is the co-founder and CEO of Qloo.