3 questions marketers should ask before investing in AI

The advent of artificial intelligence (AI) might have many marketing execs rolling their eyes and saying “Not again!”

As they well know, too often, marketing decisions come down to a guessing game. Marketing automation and analytics software can help, but the testing and measuring tools they offer only generate insights after the fact. And then there’s the meta-decision: How do you know which marketing technology will actually deliver actionable results?

Artificial intelligence technology can improve your decision-making and budgeting abilities. The question is where to apply AI and how. You don’t want to fall into the same guessing game about which areas of your marketing tech stack would most benefit from the application of AI. That just moves the guesswork from one area to a different, even harder to understand one. Suddenly you’ve got a meta-meta-problem.

Given the range of ways AI can help your marketing efforts, the natural question is: How do you make decisions about how to deploy it? Bloomberg Beta’s latest machine intelligence landscape charts hundreds of companies across 8 major categories and dozens of subcategories. The ways you can deploy AI in your marketing organization are numberless. How is a CMO supposed to make an informed decision about a technology sector that is literally founded on high-level mathematics?

The key is to have a robust framework for evaluating areas of your business that could be improved by the application of AI technology.

Here’s how I suggest evaluating where to apply AI.

1. Which business processes have the most waste?

Marketers have been familiar with waste from the days of John Wanamaker in the early 20th century, who joked about how half of the money he spent on advertising was wasted — he just didn’t know which half. Unfortunately, 100 years on, there’s still tremendous waste in the marketing pipeline. Millions of people may visit your website, but only a few thousand will enter their email addresses and become qualified leads. From there, even more will drop off in the progression to becoming true opportunities: getting in your sales pipeline and then becoming a paying customer. Only 0.3 percent of qualified leads wind up becoming customers — a ratio of 250:1. The other 249, in other words, are a waste of your effort. If you can spend less time on leads that are unlikely to convert and focus your efforts on the 0.3 percent most promising, you’ve just eliminated a lot of organizational waste.

In general, the higher the percentage of waste there is in a process, the more machines can help. And the further up the funnel you go, generally the more waste there is. So AI can be particularly helpful in focusing top-of-funnel spending, such as banner advertising and email campaigns. Don’t be drawn in by marketing activities that have a low dollar expenditure but reach a lot of people, such as email marketing. There’s a cost to poorly targeted, poorly customized email too, even if you’re only paying pennies per thousand for delivery: It hurts your brand, and, ultimately, the deliverability of that email. Even with messaging, it makes sense to use AI to generate more personalized, more effective content, reducing the waste in your email campaigns.

On the other hand, marketing to an existing customer tends to be a low-waste operation. Compared to the universe of all prospects, you have relatively few customers, and you already have a relationship with them. If your product is good and you’re taking care of your customers, you probably need AI less in this area.

High-waste areas where AI can deliver a lot of results include display advertising, website traffic, email marketing, and events.

2. What will generate economic surplus?

Economic surplus is a simple concept: What investments can you make that will generate more revenue than the cost of the investment? With many marketing decisions, it can be hard to know for sure ahead of time. But AI can assist you here by applying data modeling and predictive analytics at a much larger scale than any human could manage.

For example, imagine that you’re trying to get more users to fill out a form on your website in order to increase the conversion of site visitors to qualified leads. The traditional web marketing approach is to A/B test different versions of the form, tweaking buttons, colors, layout, and copy, until you arrive at the optimal version that converts the most site visitors. A machine learning system is not limited to testing the form itself — it can examine every possible path a visitor could take through your website, figuring out which paths lead to the most conversions when that visitor finally does reach the path. In fact, an AI system can figure out the optimal path for every visitor, adjusting elements of the site in real time to increase the chances of a successful conversion.

3. What unique data do you have?

Artificial intelligence is not magic, and it doesn’t operate in a vacuum. To be effective, AI systems need data — lots of it. And the more unique your dataset is, the more likely you are to be able to pull interesting and effective insights out of it. Also ask yourself: How good is this data? Is it high-quality, vetted data on customers? Or is the data polluted with lots of noise? Any area of your business where you have a good amount of high-quality data is an area where you should investigate applying AI.

If you don’t have a lot of reasonably clean data in a given area (such as event marketing, for instance), it’s probably best to aim your AI efforts elsewhere for now.

Taken together, these three questions can help you focus your technology efforts on applying AI where it will make the most difference. AI technologies for marketers are well-suited to creating new economic surplus opportunities and eliminating waste, and they operate best when there is a large quantity of unique data. Focus your energies accordingly.