As more and more vendors sell machine learning solutions or “AI-enhanced” features, CMOs are under increasing pressure to not only learn about AI’s capabilities, but sort through the many lingo-heavy sales pitches that cross their desk. “Aren’t we doing machine learning already? Isn’t that what my DMP does?” (they may ask).
AI is the shiny new feature that’s in every sales deck across marketing, media and ecommerce. Its promises are substantial, but the learning curve can be steep for the uninitiated.
Steve Irvine, founder and CEO of Integrate.AI, addressed some of these issues at strategy‘s recent Marketing Evolution: C-Suite Summit. As a former senior Facebook exec who led the platform’s interactions with brands, he’s seen how companies struggle to understand and integrate new technologies into legacy business operations. Through his on-stage interview (conducted by Mirum president Mitch Joel), Irvine set marketers’ expectations as to how to start building their brands’ AI capabilities.
What’s the definition of AI vs. Machine Learning?
“There’s the technical definition of AI, which is the study of computer systems replicating human intelligence – learning, vision, natural language processing. The other way I think of AI is ambiguous: it’s everything computers can’t do until they can… Machine learning is easiest to think of as a subset of AI. The terms are used interchangeably, but they do have different definitions. Machine learning is computer systems learning without humans having to go in and program them.”
What should marketers be asking potential AI partners?
“Mistake number one tends to be that we over-complicate short-term opportunities. If your number-one project in AI is to build a robot that takes over your store, emulates Alexa and responds to people in real-time to change your business, that’s a bad first step. The common fallacy is wanting to put a body to AI. The conversation becomes about robotics, self-driving cars or other substantiations. AI is just the software.
“The low-hanging fruit is taking your data, understanding where it would be helpful to add to make business better, and then doing it.
“I don’t think anything in AI should distract from your [business] priorities. It’s likely that your biggest asset is in your data, although many companies may not have good data, or the right kind of data. You want data that has clear outcomes that you care about and some sort of path to that outcome. The easy example is ecommerce. You’ve got a lot of info on the path to purchase where the purchase is clear and connected back to the individual. You can train an algorithm with that kind of data to build predictions about what people will want in the future.”
How is the promise of AI any different than what ‘big data’ promised in past years?
“There’s a hierarchy in the data science world. Big data just means you have a lot of data; it’s the most abused term of all time. Then there’s analytics – doing some arithmetic, adding stuff, multiplying it, that’s really all you’re doing. Then there’s data science, which gets into modelling and trends over time, trends that are easier to map.
“Machine learning is where you get into a sophisticated area. A PhD in machine learning is like a PhD in astrophysics. It’s applying really complex math to find patterns that are really difficult to find.”
Should marketers be the ones buying AI?
“Some people have religion on this. I don’t. Where this breaks down in traditional companies is where there are very defined lines around who owns the data. For many companies, IT owns the data, marketing owns the campaigns, but the P and L holder owns the budget. It gets complicated. You have to find a way to break that down. You won’t be successful unless you have some understanding of the data in your business. Whether that’s through a partnership or internal teams, it’s important to break those walls down.”
