One of the hallmarks of artificial intelligence (AI) is its ability to learn, react and respond to its environment.
AI continuously observes the changing set of stimuli (i.e. data points) surrounding it, both from the point of view of what it is programmed to know (internal) and what it is able to experience (external).
As I wrote in the first article of this series, this mechanism allows us to evaluate AI on its ability to be personal.
However, when speaking with someone on a personal level you need a deeper context, you must respond or adjust to every clue the person gives off. You must also be prescriptive in the way you advise or inform them back.
How does AI hold up against these tests?
Is AI Responsive?
Language Is Complicated
Nothing is more central to intelligence and more variable, evolving and personal than language. It is — at its core — the one human ability that enables everything we do, and conversely can disable or block everything as well.
I may tell a sales associate that I want “swim shoes,” but that doesn’t mean I'm looking for a pair of flippers. A sales associate at a high-end retailer would likely know that I was actually looking for Swims shoes.
Likewise, as businesses and marketers trying to understand and evaluate the practical applications for AI, we must start by acknowledging that nothing is precise about language.
Words, tone and context each carry different and changing meanings. We’ve already seen these issues crop up with natural-language and AI-driven chatbots, which is causing some brands to drop their chatbots due to high failure rates.
Let me go a bit deeper into tone and context to relate in terms of the metrics we all know and love [to debate].
When a customer walks into a business, we can generally tell if he or she is happy, frustrated, explorative or determined. Facial expressions, tone of voice and body language all give human intelligence inputs about how to address the scenario and respond.
Can artificial intelligence emulate this capability based on digital signals? Turns out, when customers digitally engage with a brand, they're giving plenty to go on.
As marketers, we often view increased time on site and more page views in multiple instances of single sessions as good indicators, but does that mean the customer was happy? It’s possible, but it could also mean that customers are lost and frustrated, with a high likelihood of never returning.
Think about the deeper implications of this scenario. That means all of the assumptions about linked products or content could be completely false. So, it’s up to “intelligence” to discern, both at an individual and collective level, what context is powering the correlations and assumptions between data points.
Similar to a salesperson watching a customer wandering through a store with a certain look on their face, in the digital realm it’s important to critically determine context based on subtle clues.
AI Cannot Operate in a Vacuum
Being responsive in a closed ecosystem is extremely hard. To move a digital visitor to a desired outcome requires intelligence, but intelligence cannot adjust according to what it does not know. We rarely can determine a single source of truth from data located within specific silos or within our own “walls.”
Instead, for AI to be responsive, it must understand behavior and context at a broader scale. It should either have access to or be able to incorporate data from outside the specific site or brand its powering.
Going back to my language example, the way consumers use words to find what they want is constantly changing. Certain products or services on a blog, news site or even the physical world may influence them to describe those goods or services differently every day. It’s up to AI to recognize those shifts and incorporate those new associations back into its system.
And, whether it’s through open APIs or across applications, AI should share that knowledge with other systems that handle other parts of the experience.
In addition, just as humans can make wrong assumptions, AI should constantly test itself to see if its assumptions are right.
For example, as the entire industry moves toward more 1:1 personalization, AI will increasingly be the driving force powering personalization. As it begins to offer content targeted toward an individual, that customer may exhibit behavior that they are not interested. Maybe their intent is changing, or maybe it isn’t the right customer, but it’s important that AI respond by adjusting its suggestions in near-real-time.
The underlying point is that every factor driving customer intent is constantly changing, and effective use cases for AI cannot operate in a vacuum. As businesses look to invest more in their AI initiatives, it’s important to look for those that embrace an open architecture.
Is AI Prescriptive?
Exploding Sources of Data
While much of my discussion so far has focused on the customer experience side, AI plays a role on the business user’s end as well.
This is where the human-plus-machine model reaches its paramount importance. Over the last decade, the big data movement and advancements in machine learning have allowed CMOs, CDOs and CIOs to become better strategists — not to mention work more collaboratively. It’s been able to condense ungodly amounts of information down to consumable, actionable chunks.
However, as the data sources, the ways to slice it and applications to analyze it exploded — especially in the marketing and sales landscape — we’ve arrived at a place where the ungodly is now the unmanageable.
In short, it’s not enough to compete in the current market. And I think would this is where a lot of the interest (and hype) around AI has derived. What once freed up human resources for higher-level strategy or creativity now leaves way too many insights to sift through and conclusions to act upon.
Increased Pressure for Late Adopters
Some of AI's early adopters have put their competitors only a year or so behind at a tremendous disadvantage. This makes an enterprise’s AI strategy that much more immediate and complex.
It’s not enough for “intelligent” applications to identify issues and opportunities: it now must tell the business user what to do for the business user’s specific role while including the most likely outcome.
Artificial intelligence should be a prioritization technology, similar to a business intelligence unit analyzing and advising upon request and in real time. Those driving digital experiences in a truly personal manner need to throw out the rule books and spend time doing what humans do best: tuning and testing the experience based on intuition and non-analytical, gut-driven risk-taking.
Striking a Balance Between Humans and Machines
Does AI pose a threat to industry professionals? The simple answer is “no.”
As the content and digital experience markets move toward more role-based, AI-driven applications, professionals shouldn't worry about large-scale job obsolescence.
Every time an industry reaches a certain inflection point, it seems like some see it (and fear it) as an opportunity to consolidate human capital. We've seen tremendous advances in automation over the past 20-plus years, and each time careers have adapted to become more specialized.
There are some things that humans just understand and do better than machines — at least for the far-reaching foreseeable future.
However, until the time when the apocalyptic rise-of-the-machines happens, we should continue to see AI as an opportunity to improve business performance, while refining our ability to apply it. In the content, commerce and digital experience world, that means we should consider and evaluate AI by questioning its ability to be personal, responsive and prescriptive.
Editor's Note: Read the first article in this two-part series here.