Webinar: Machine Learning To Increase Human Understanding



On April 13, 2017, Assaf Baciu — Persado’s very own co-founder and head of product — delivered a webinar presentation hosted by BrightTalk on the topic of using machine learning technologies to enhance human understanding. In particular, Baciu discussed how Fortune 1000 brands leverage machine learning to create intimate emotional bonds with their audiences.

There is a misconception that infusing automation into the customer experience will make it less human, and thus less relatable, to consumers. On the contrary, companies are pioneering AI technologies to enable them to gain a deeper understanding of the people they serve, resulting in experiences that are more relevant, personal, emotional, and ultimately rewarding — for customer and company alike.

One point that Baciu stressed in his presentation is the relative lack of utility of traditional persona-based marketing. While knowing income, region, marital status, or “was browsing mattresses recently” might be useful for targeting, it gives no indication of how to talk to someone.


If you were to meet the above “person” in real life, would you know, based on the available information, exactly how to talk to them before ever seeing their face? You wouldn’t. It’s impossible. You couldn’t plan for it.

But once you met them — saw their face, clothes, body language, etc., you’d pick up on social cues and have a good idea of how to relate to them. When you talk to someone, you constantly process signals: vocal tone, facial expressions, body language, et al. For most people, this process is almost completely involuntary.

Humans detect social cues and react. It’s really the most efficient way to be liked, and being liked has always come in handy for survival.

Companies try to understand in this same way, but they’re limited. The information we get from digital interactions isn’t rich enough to give us the kind of information a face-to-face interaction might give.

A.I. can do the same thing. But A.I. uses more data, across millions of people. And this ability is transforming business and marketing communication as we know it. So instead of bland facts, you get an emotional profile that gives you even more information than you’d get sitting across from that same person.


Companies talk to millions of consumers at any given point. That’s the challenge. How do you adjust the message so it’s on point, on brand, and uses language that speaks to a specific person?

This was really the crux of Baciu’s presentation. There’s no difference in how you’d address someone whose annual salary is $50k versus another with $150k. But you would know how to talk to someone who consistently reacted to marketing that elicited excitement versus someone who never, not once, clicked on a message that conveyed excitement.

The new era of marketing will focus on emotion. It’s what marketing has always been about. Remember the famous Mad Men episode “The Wheel” from Season 1?

But “emotional targeting,” the type with which Don was so adept, is not as monolithic as Mad Men makes it seem. There’s no “right emotion” to use for a product. But there is the right emotion to use when addressing a single person.


As Don said at the beginning of the above scene:

Technology is the glittering lure, but there’s the rare occasion that the public can be engaged on a level beyond flash. They have a sentimental bond with your product.

The same can be said of the fleet of A.I. solutions storming the market. Can this flashy technology actually help foster an emotional bond with your brand?

This is why Baciu closed his presentation with three criteria for assessing A.I. for business.

1) Always evaluate your A.I. solution’s decisions by comparing the output to the original hypothesis. More and more machine-learning algorithms and AI systems automate our society. It’s a big leap forward. If an ML decides if I get a mortgage, what’s my curriculum, what ads I see — then I need to compare my hypothesis with the outcome

2) Your A.I. should tell you what it is missing, what it needs, to make decisions. It may make a decision even with corrupt data or missing data. There must be a failsafe mechanism so it lets you know if you don’t have enough data So in marketing — are your results significant statistically?

3) If there is a conflict with the first two laws, put your A.I. to rest. You go back to the beginning with a different system that will deliver on its objective.

The real benefit comes when a machine-learning solution proves itself:

A machine can interact with millions of consumers at the same time. It can crunch all the response data — the “signal” — and use that as actionable insight. It can remember how every consumer has responded. Over time, it can tune itself to individuals. We can analyze massive amounts of data with machine-learning that humans may overlook.

In the case of Persado:

What was the structure of the message? What was the verb tense? Was there an exclamation point? These subtle differences are difficult to pick up on en masse, but the machine can respond immediately and give back a better experience the next time. That’s the benefit.

In case you missed this wonderful presentation, first, shame on you. And second, check out the recording below.