What Does AI Mean for the Future of Marketing?
Crystal McFerran, CMO | The 20
The next big leap for artificial intelligence has “been around the corner” for decades. It’s not that the technology hasn’t been advancing, it’s just that the more we learn the less we know about what to do to really raise the bar. We’ve been waiting for the AI equivalent of the transistor, but nothing seems to really nail it to bring us into the Star Trek utopia singularity we’ve been promised (or at least make our jobs that much easier).
While it’s too early to really determine whether using a light based system over a standard electrical one will be that change, it will do what most AI advances do and open new doors which push everything forward. From the preliminary results we’re talking about an order of magnitude or more of improvement, but will this scale with current AI is the question? I can’t claim to know enough to give an answer, but what I do know is that the prospect of more accessible AI opens the door to many life changing prospects from career and on.
AI being easily applicable in marketing would be game changing for the field. The old way of marketing has had a rougher year than most of us, pushing digital marketing to the front. Despite the chaos around the world, marketing is seeing an almost Renaissance level of change and revitalization. The core of classic marketing is being rebirthed into the current era. It’s not much of a silver lining, but I’d rather see the drop in the glass than say it’s dry.
How Will AI Change Marketing?
Marketing has largely moved from physical networking to digital marketing. Your business card is something nice to have, but not a necessity anymore either in many circles. Going to a conference can help you network and forge deals, but it’s not the penultimate solution it was previously sold as. (That isn’t to say the traditional methods are dead, they have become an element rather than a full solution). It’s cheap and easy to reach out to basically any market anywhere in the world with the globalization of the internet and the collapse in advertising costs.
You aren’t constrained to a single market or markets you can physically reach, you’re constrained to your solution solving a specific problem. A plumber in Denver isn’t going to fix anything in Florida, but that doesn’t mean that digital marketing isn’t essential. Even local businesses are found mostly through search results and digital ads.
The problem your solution answers is constrained by region, but your outreach isn’t. This is both a blessing and a curse. The extra reach is a byproduct of your effort, not really a cost when done right, but that just means there are that many extra results in each search. It’s cheaper than ever to market which means everyone is doing it, you have to learn to do it better.
The big problem is that we have so much more data than we have ways to make sense of it all. People have gotten used to more and more personal outreaches (while expecting fewer in person and phone outreaches), and expect things to be more and more tailored to them while wanting to volunteer less and less. You have to read between the lines, but you need to know which lines matter too. Knowing where to go next and how much to use for any part of a campaign just gets increasingly complicated with the plethora of variables to track. There are also things which are repetitive but painfully manual due to technical limits. These are the problems AI in marketing will (eventually) solve.
We won’t start at AI, but we’ll work there as the technology becomes financially feasible. Amazon, Microsoft, and others have each released machine learning and AI frameworks, but they can be cost prohibitive. Frameworks like Torch exist, but they require you to understand more than just the recipes. The big barriers to entry are the cost in terms of both skills and platforms in order to really get moving. You need the right people who can create, condense, coalesce, and coordinate every piece of relevant data into something which can be further processed into something meaningful. The Amazon of today for that level of AI and machine learning (ML) will be the one-person marketing hustler of tomorrow once we find the right catalyst.
Automating Data Science
Virtually no one in marketing or otherwise has a problem getting data, they have a problem in getting the right data. You can either build data as necessary which is time consuming and slow, or try to get a massive dump of data and cut it down. These processes can be combined in different ratios and different stages as the secret sauce of your marketing program, but it’s still the same two ideas at play.
When should you acquire bulk data and scale down, and when should you slowly build a list? Just how should you cut out unnecessary bits and just how should you add individual entries? There isn’t a universally correct answer, you have to have some kind of metric for each effect, a way to measure changes from your action(s), and the knowledge and wisdom to make it work.
Ideally, AI will eventually take this sort of task over. You guide it based on what your campaign aims to do, and you offload the pain to a machine which won’t make mistakes (unless misprogrammed to do so). The seeds of change already exist in various analytic systems and similar which help collate and break down data. When it gets economic to do so, we’ll see a huge jump in platforms working in machine learning and beyond in some form. This process will either target increasing efficiency, decreasing inefficiency, or a balance.
Campaign Personalization
Once we have a bunch of data, we need something to do with it. If a specific campaign hits on 75% of the metrics for a subset of an audience, but the average take from the same campaign is 50%, how do we only target the more profitable side with that piece? How do we know what kind of campaign to build and balance? Your data is working to help you identify how to bring in clients, how to keep them happy, and how to get them to your sales team to close the deal. It also helps you figure out how to retain who you’ve already pulled in (if you use it right).
All of these sorts of things exist, but they’re buried in the data and metrics. It’s beyond what the average person can make sense of all the time. You need to have a way to reduce the data to something a person can cognitively process, then find a way to make use of that data. If you find that you get a nice Pareto distribution of a specific group from your data, you can personalize campaigns for them.
That can be cost prohibitive now because by the time you run the numbers, the campaign is usually already dusty with more traditional methods and more diverse campaigns. AI will make crunching the numbers for that something far easier. What are the unifying factors between various groups and what divides them? You could make 10 variations of each section and a program could reassemble them easily, but then the content looks fake without multiple people to proof and test. AI will provide a way to accommodate these sorts of changes without having to reinvent the wheel at each step.
You have to keep track of which people are in which group, and which campaign hits which group or groups. Identities change and it’s easy to miss subtle changes in a person the traditional way. A machine doesn’t make mistakes, but you can and will. Most of these ideas exist in some consumable form, but they just aren’t as mature as they could be. As automation accelerates, we’ll see more and more personalization with a cut to budgets due to the extra efficiency.
Resource Utilization
Part of increasing efficiency and reducing inefficiency is getting the most out of your resource utilization. If you know what is important (and most important) for a campaign, it can be easy to refocus your resource utilization in a way that makes sense. You don’t put your best graphic designer on a low importance task if you can help it.
If you know what matters, you know what you can target. This doesn’t need to be some Orwellian level robot overlord, but you want to be able to more efficiently use your resources. Where are things falling through the cracks and what is missing its mark, without being creepy?
HR, managers, and various leads currently try to balance this role, but there is so much more to it which takes way more people than it feels it should to get an underwhelming level of efficiency (compared to what is possible). You need someone making sure things function correctly, but you also need management to translate tasks between one group to another, and you also need oversight that the process hits the requisite checkboxes.
The right analytics, metrics, data, and algorithm could make this process more efficient. We probably don’t want to take the humanity out of the process anytime soon (if ever), but a lot of wasted labor could be reduced so that people can accomplish more important tasks rather than communicating the same things differently a dozen times a day.
Technical Limitations
Whether or not light based TPUs will bring us to the singularity remains to be seen, but each development is a step towards a more automated future. As technology progresses and our interactions with it do too, we’ll continue to see more and more growth in the applications of advanced data science with marketing and everything. AI is arguably just advanced data science with the right math to make things work. It sounds simple, but it’s like saying marketing is just sending the right communication. The devil is in the details for each part of the process.
Technical limitations, including scale and cost are what really hold us back from further use of AI in marketing and beyond. As the technology matures more and more, we’ll see data science and ML take off. It’s begun in some places, but it’s a bit like the Dot Com era, a Wild West of technology complete with the digital snake oil salesmen. Some supposed AI is just a weak ML principle with lot of “if” statements to make it work. It’s a black box by necessity and by design.
As the veil is lifted and it becomes easier to more scientifically approach the application of ML and AI in marketing, we will see a shift at even the lowest level to more integrated data science. Moving the apex moves the base of what is acceptable. The bar will get higher and higher until we hit the true singularity for AI. Until then, we’ll see advanced data science grow and grow in application in modern marketing from big to small. As more and more data can be produced and processed, we’ll see more and more automation of actions on this data. The singularity is coming, but that doesn’t mean you can’t grow from it first.