What Does AI Need to Succeed for Marketers: Human Intelligence

Charlie Neer

July 9, 2019

Artificial Intelligence is widely touted as the solution to all business challenges, especially the ones faced by marketers seeking to build their top-funnel pipeline. AI will reach and engage high-value prospects, Marketers are promised. If so, why are so many marketers disappointed by the results? It’s time we stop thinking about Artificial Intelligence as the panacea for all marketing challenges. If we keep on in this vein, marketers will rightly think of it as snake oil – full of promises it can’t possibly deliver. This is a shame because AI has a lot of potential, and limitations. The truth is, with Human intelligence involved, AI works best.

How can Human Involvement Improve AI Performance?

Humans are essential for two key reasons. The first is to unlock the efficiency of machines. The second reason is that only humans can bring the personal, warm, and welcoming aspect that brands need to deliver successful campaigns.

To be sure, there is a long list of tasks that AI is especially good at. It excels at Aggregating, Segregating, and Storing data in effective ways. It’s also pretty useful for making initial assumptions based on scale. For instance, unsupervised Machine Learning can crunch through massive datasets to identify statistically relevant connections that reveal useful and actionable insights for marketers. However, we should never forget that those insights are assumptions that a human must verify.

Why AI needs Human Intelligence to succeed?

We all know that Siri is a sophisticated AI platform, but we also know that it requires a considerable amount of corrections and instructions by the individual device user to work properly. This isn’t bad; it’s simply the reality of AI.

In contrast, there are things that AI isn’t very good at, and if not corrected, can lead to embarrassingly poor brand experiences. Topping the list – AI cannot eliminate any biases that are unknowingly built into an algorithm. This happened to Amazon when it launched an AI tool to help screen employment candidates. As it turned out, the algorithm had developed a bias for male candidates, and ceased to forward resumes of female candidates for consideration. The effort had to be abandoned; worse, Amazon undoubtedly missed out on some excellent candidates.

As Alexandria Ocasio-Cortez recently said of AI, “They’re just automated assumptions. And, if you don’t fix the bias, then you are just automating the bias.”

We see this bias every day in our News Feeds, which show us only stories that reflect our political bent and harden our beliefs. We are becoming more polarized because we are rarely exposing ourselves to stories that offer opinions that differ from our own.

Nor can AI understand the complex nuances of human emotion, which Burger King proved in the most delightful and hilarious ways when it allowed AI to develop its TV commercials. If you haven’t seen them be sure to check them out; they’re side-splittingly funny, if not effective at selling products. Branding is all about tapping into deep human emotions that prompt a consumer to act; AI is useless in understanding why or how emotional connections occur.

In all these cases, human intelligence would immediately spot when and how AI got things wrong and steer it back in the right direction. Human intelligence would begin to suspect a problem after seeing just a handful of male resumes; by the time reaching two dozens, human intelligence will pull the plug on the tool. This is why AI will never succeed without human intelligence.

Marketers who believe they can sit back and let a machine decide who their prospects mostly likely are and which messages they should see are in for a rude awakening. Data scientists will always be needed to assess a marketer’s business needs, and identify data sources that are appropriate to uncover answers (e.g. which data sources will help windshield manufacturers understand the factors and conditions that lead to broken windshields and thus, the right time to target consumers).

When companies leverage their best resource [people] with customized Machine Learning assistance [AI], true value can be unlocked and leveraged.

AI will always need data engineers to translate complex APIs to a common language so they can be deployed in a campaign. For example, when students head back to school, they are more likely to buy new furniture or dorm room needs. Therefore, when students are looking for a new school swag, the manufacturer should update their strategy and pay more to reach those consumers within a college town or area. It is only at this point that AI can take over and automate decision-making. This is exactly how the industry has set up programmatic advertising campaigns.

Conclusion

Make no mistake about it, when AI is focusing on specific aspects of a campaign it will add mind-boggling efficiency which will lead to rich rewards for the marketer. But as is with the case of programmatic, human intelligence will still need to keep an eye on the algorithms and optimize it based on results.Without it, AI applications will always carry a risk that they’ll offend, amplify existing biases, and deliver the results marketers would prefer to avoid.

Charlie Neer

Charlie Neer has a proven track record of taking complexities from the ever-evolving digital advertising space and simplifying them into solutions for companies and their clients. As SVP of Revenue at MiQ, Charlie’s dedication to building a presence across the Midwest and West Coast regions has led to unprecedented growth. Previously, Charlie has held positions overseeing strategic partnerships and revenue at Adaptly and spearheading the technology development of a DSP at Click District. From early on, Charlie has always been fascinated by two things - technology and business. Through the course of his career, he’s held positions that have allowed him to play in the cutting edge of both of those spaces. He is a consultative seller and leader that puts a high value on listening to colleagues, prospects, and clients.