The Marketing AI Institute held its annual MAICON conference this month. I dropped into a few sessions hoping to learn more about Artificial Intelligence (AI).
Marketer plus machine
The event kicked off with a keynote from the Marketing AI Institute founder, Paul Roetzer. He shared an eye-opening stat from McKinsey Global that estimates a 6 trillion dollar impact of AI and other analytics on marketing and sales.
Although we understand the impact of AI, and there is plenty of interest and awareness, we still have a long way to go, Roetzer said.
We haven't reached the tipping point.
The institute performs an annual survey to understand where companies are with their understanding and use of AI. This year's results show that most people classify themselves as beginner or intermediate in their AI knowledge.
But here's the thing Roetzer said that hit home - you don't need to learn AI from the perspective of being able to code and build models and algorithms. I suspect that's what scares most marketers - that they need to have this intimate understanding of the inner workings of AI models.
What you need to do, Roetzer said, is use the applications and let the AI help you use them better. So instead of the marketer having to learn how to build AI, they allow the machine to learn and improve continually. The marketer's job is to monitor and measure its performance. As a result, he said:
AI is just smarter marketing technology that builds smarter businesses. It's not about how advanced the AI is. It's about how much efficiency and performance can improve through intelligent automation of processes and tasks.
But not all AI is created equal, and using it does not always provide better outcomes, Roetzer cautioned. There's a big disconnect in the market today, and a lot of it comes down to the vendors. Sometimes the product isn't mature enough. Other times, the AI capabilities are overhyped in branding. There is also sometimes a lack of education for marketing and sales teams on precisely what the product does or how it works.
Organizations have a responsibility to ask the right questions when they evaluate AI-enabled marketing and sales technology.
Assessing the level of AI in a marketing platform
To help organizations assess AI-enabled marketing technology, the Marketing AI Institute developed an assessment guide called the Marketer to Machine ScaleTM. It's a scale from level 0 to level 4 that classifies the amount of intelligent automation a solution provides at the use case level.
The use case is critical here. Roetzer said you can't evaluate an entire platform. Instead, you have to look at the technology in terms of how it supports a use case. Four variables together determine the level a solution fits on this scale:
- What information does the marketer need to provide for the machine to perform the task?
- How much training, monitoring, and intervention is required.
- How reliant is the machine on the marketer to complete the task?
- How does the machine learn and improve?
Roetzer said that most solutions we buy today sit at a Level 1 (mostly marketer) or 2 (marketing plus machine) and that Level 3 (mostly machine) is possible. But Level 4 (all machine) does not exist in marketing today.
Natural language processing and generation - a marketer's perspective
Christopher Penn, co-founder of TrustInsights, said that marketers care about natural language processing (NLP) and generation for an important reason - marketers have a "treasure-trove" of data, and they can't unlock its value. At its core, NLP turns words into numbers so that machines can analyze them and provide us with insights into our customers, he said. "It's statistics and probability." It's not understanding or human.
Penn's session was an introduction to NLP in marketing, and he shared a set of use cases from least complex to very complex. For example, word clouds use NLP, but it's basically counting how many times a word appears in a dataset. So the word that appears most is the prominent one in the word cloud.
Another example is a good one for marketers who want to understand their audience better - analyze the words in their Twitter followers' bios. Then, a marketer could use this information to create the best content for their audience.
The most complex example Penn shared was topic modeling, which is understanding how words relate to each other. So you could ask your customers open-ended surveys and apply NLP to analyze the responses by looking at words and phrases that occur close to each other to determine key topics and ideas.
As you bring in more and more data, Penn said, your modeling continues to adapt, which means you will see when or how things change over time.
Marketing software applies NLP, including Talk Walker, MarketMuse, Google Ads, and SEO tools. But you can also build solutions internally if you have the time and money to do it. Penn suggested that if your strategy is to make marketing more efficient, then buy what you need. On the other hand, if your strategy includes NLP as part of your secret sauce, then build your solutions.
He also said that you should start with a platform. First, you need to define your user story and what you're trying to do. Then make sure you have the people necessary to accomplish that purpose: subject matter experts who can review the results, technical resources to operate the tools, and statistician/data scientist who will help you understand if the data is meaningful or useful. From there, you design your processes and select a platform, and monitor the performance.
Growing smarter with AI
"A little bit of AI can go a long way." Your entire marketing department doesn't need AI, and every piece of software you use doesn't need to be AI-enabled. There are many use cases where it can improve processes but instead of trying to do everything, think about the use cases that could benefit the most.
Roetzer ended his keynote with a five-step process to grow smarter with AI:
- Understand what AI is first.
- Pick your use cases - think narrow scope, high probability of success, quick wins (he also suggested to stack your projects).
- Research potential vendors.
- Calculate the expected value in terms of efficiency and performance.
- Prepare your team with the right training and education.
I wasn't able to attend all the sessions of the two-day conference, but there were plenty of interesting discussions, like the keynote with Karen Hao, the Senior AI Editor at MIT Technology Review. She talked about responsible AI and what's happening at the big companies like Facebook and Google.
My biggest takeaway from her conversation with Roetzer is that organizations have a responsibility to leverage artificial intelligence ethically. There are many ways we can adopt it in our marketing and other programs, but what we do should give our customers better experiences. Brands also need to be open and transparent about how we leverage AI. Unfortunately, today, we don't always see that as the case, and consumers are often suffering as a result.
Also see my MAICON piece on keynoter Mathew Sweezey, Meet the post-AI consumer, says Salesforce's Mathew Sweezey.