Marketing is becoming increasingly automated, and before long there may be little need for lights in the marketing department. Humans will still oversee these automated efforts, but there will probably be substantially fewer of them.

It’s not that the goals of marketing have changed, but the means to achieve them. Companies are typically trying to define and target specific customers or segments, but if there are thousands or millions of customers, automation is needed to get to that level of detail. Companies are also trying to customize the experience of the customer, and that also requires automation-driven detail. Automation also helps to deliver value across omni-channel customer relationships, and to ensure effective communications at all customer touchpoints. Finally, automation lets companies make decisions with similar criteria across the digital and analog marketing worlds.

Automated marketing applications typically have a few things in common. They support repetitive decisions based on data, each automated decision has low monetary value (though in total they add up to large numbers), and the decisions mostly involve digital content and channels or online promotions. Of course, almost all content is becoming digitized, so it makes for a pretty big category. And even non-digital marketing is increasingly automated.

Digital advertising is already highly automated today. So-called “programmatic buying” matches a company’s digital ads with publishers while determining the price to be paid for the ad (often with an auction) and some degree of targeting to the viewer. This decision is typically made within a few milliseconds, so it would not be feasible for a human to make it. Some ad agencies have programmatic marketing capabilities, and their advertisers turn the management of the process over to them. Increasing numbers of companies, however—including Procter & Gamble, Allstate Insurance, American Express, Unilever, and Mondelez International—are controlling their own programmatic buying, and are moving to doing all digital ad buying that way.

As Bonin Bough, VP of Global Media and Consumer Engagement at Mondelez, put it at a conference:

I don’t see a reason why brands wouldn’t go 100% programmatic, or something close to it…Why would I not buy media in a way that I can get data back and use it to make smart decisions?

Two other highly digital domains, website operation and optimization, are increasingly automated, and include several different types of activities. Search engine optimization (SEO) tools evaluate and recommend links, search terms, site structure, and other attributes to improve search rankings without any human intervention. A/B testing compares different versions of websites to optimize clickthroughs, conversions, and other metrics, and is largely automated among more sophisticated practitioners.

Even the production of narratives about website viewing patterns is automated. National Positions, a company that does online activity reporting for other companies, uses Wordsmith for Marketing software from Automated Insights to generate automated reports about website and search engine activity using Google Analytics data. The reports not only describe weekly and monthly results in written English, but also provide some possible explanations for changes in activity.

Here’s a sample of the kind of text that gets generated automatically; it’s about the monthly web analytics results for a real estate company:

In February, Emerson Real Estate had a significant gain in visits relative to January, as they rose 24.04% to 792,385. Visits were up across three channels, led by Direct which added 116,858. Along with a big rise from January, traffic to the site was up 60.66% compared to February of last year. The jump in visits was correlated with gains in two referring sites, one search landing page ( /corporate/home.aspx), one paid keyword (chapel hill real estate), seven campaigns, and three social networks.

Granted that it doesn’t read like a Stephen King novel, but it does what automated text generation generally does well, which is to write clearly (and cheaply) about data-intensive topics.

“Marketing automation” sounds like an umbrella term for all such activities, but it primarily involves software to automate a variety of activities other than digital advertising or website management. They include targeted email marketing, automated creation of landing pages, customer segmentation, lead filtering and scoring, and some social marketing. Marketing automation in this context is a well-understood software category, with a variety of vendors in the space.

Some organizations develop this capability on their own and make it a high-volume marketing activity. For example, at Wayfair.com, a large online shopping site for home goods, a primary focus for automation is personalized emails to customers. Ed Macri, the company’s Senior Vice President of Marketing and Analytics (a title that tells a story in itself), comments that this type of marketing automation is well-institutionalized and effective:

Each day we send out over a million different versions of our marketing email. It’s personalized based on customers’ past purchases and most importantly by their recent activity on our site. We learn categories of interest as well as style preferences. For example, I spent a Sunday morning shopping for bathroom fixtures on our site as I was planning to remodel a bathroom. On Monday, the subject line of the email I got from Wayfair.com was “Start your day refreshed with Kohler, Delta, and more.” It was a perfect fit and I was thrilled to get the email.

There are a variety of non-digital marketing activities that are also increasingly automated. Stay tuned for more about those, and what this all means for marketing departments in the present and future.

Originally published in WSJ’s CIO Journal