Behind the Curtain: How Algorithms Impact Performance
We recently participated in a “Behind the Curtain” panel hosted by the Dallas Fort Worth Interactive Marketing Association (DFWIMA). The discussion focused on the viability of current state algorithms, as well as how can we approach, plan and implement AI appropriately for our organizations and clients. As part of T3’s internal task force on this topic, I was excited to share how we are growing this expertise at our agency and learning about how others are doing the same. While nuances definitely existed, a few major themes emerged:
Experiment: Just Try It
Last year saw some major AI-driven technologies such as chatbots, voice-enabled smart speakers and self-driving cars get a lot of attention. Some of the projects to hit the market were successful and some were failures. In all cases, there was a lot to learn from seeing them in action. For me, seeing chatbots fail was as encouraging as seeing them meet our expectations because it means we’re testing the technology to see what it can and can’t do. We’re figuring out the platform and building it instead of just talking about it.
This “learning by doing” mentality has become critical to our process at T3, and it’s led to some success (see our Trump and Dump bot as an example). We’re not alone in this approach either. Other panelists at DFWIMA were taking the same approach, and templatized services like Amazon Blueprint and TensorFlow are making it even easier for others to do the same.
As a result of our test and learn approach, we’re getting a better understanding of what users want and the expectations they have (regardless of how unforgiving those expectations might be). In addition, we now have a much better grasp on the data we have or don’t have when it comes to creating the right experience and evaluating the results of any given project—all necessary factors before we can truly let go of the reins on the true capabilities of this technology.
Explore: Use AI As a Way to Identify Gaps
Another key observation we’ve discovered in our work at T3 is that AI is not always the solution. Other people on our panel at DFWIMA said the same thing. While AI has incredible capabilities even in the narrow focus we see today, the inputs you need to feed it may not exist.
Since AI requires a specific vision for what and where you’d like its assistance, you can’t simply plug in AI to mask a gap in knowledge, customer experience or operations. Instead you will need to start determining what data points can close these gaps and what tools you can use to start capturing that information. While not directly an AI project, you now have a valuable roadmap of how you can take advantage of this technology.
For those leading these efforts and looking for results, it’s your job to point AI in the right direction and leverage the right channels, just like any other platform.
Empower: New Ways of Consuming & Collaborating
The final major takeaway on our DFWIMA panel was that AI is going to have a major impact on what we produce and how we work as agency teams, marketers and humans. Similar to major technological revolutions like the printing press and mobile phone, AI will transform how we consume content and connect with each other and our machines. Already, AI exists in our lives and continues to be put to use in intriguing new ways. An obvious example is the Spotify algorithms that curate weekly playlists for each listener. This algorithm pushed the platform from great music on-demand to personalized playlists made just for you. It augments the mixtape-loving friend you’d always go to for new songs. Or take a look at the The Moment, a new movie that the audience watches while wearing neural net headsets, allowing them to change elements of the movie based on their thoughts. Just consider what happens when the movie turns into a fully immersive virtual world that you can interact with real time, or become a lead character in, like a particular Black Mirror TV episode. AI is the mechanism fueling all of these personalized or predictive experiences, which leaves us as the algorithmic creators.
The question we face now is how do we solve and create for these new AI-fueled experiences? The consensus on the panel was to take a multi-disciplinary approach. AI requires an understanding of the business, the consumer and the technology all at once. A strategy for this technology cannot happen without recognizing its limitations, similar to how product designers have to consider the nuances of a brand before creating something new. The promising thing is that we are already seeing new disciplines emerging from in-house teams that are teaching themselves, and then testing out how to bring together the knowledge for the right user experience.