Welcome to DTC Future Proof, our monthly editorial series in partnership with Chris Rempel at the Lazy Marketer, where we look at the biggest opportunities and risks that will shape and disrupt the future of marketing.
Don’t forget to let us know what you think of this piece by replying to this email.
To the marketers of the ‘80s and ’90s, the prospect of AI truly becoming a marketing super weapon in the future seemed fascinating – even if somewhat unbelievable.
It was a future where “intelligent agents” could form personalized relationships with customers at an unlimited scale, where algorithms would shoulder the lion’s share of marketing optimization, and where strategy and ideation could be supercharged by tapping into an omniscient digital sage.
As thoughtful as many of these projections turned out to be (for example, as far back as 1985, experts understood the use-case for natural language processing), in many other ways, our marketing “ancestors” couldn’t see past the linear extrapolation of their own biases.
Let’s anchor this with some concrete examples, courtesy of this 1996 Harvard Business Review article on the emergence of “interactive marketing”, written back when the internet’s total population was just 35 Million users... Versus today’s 5 Billion.
And this was way before anyone had any notion of just how much AI would play a role in the “interactive marketing” that would later materialize.
The article now serves as a time capsule – complete with postulations and predictions about online marketing from contemporary leaders in the tech industry, government, and academia.
Let’s see how well they’ve held up, 26 years later...
Here’s what they got right:
And here’s what missed the mark:
With the benefit of hindsight, it’s easy to look back and see why things panned out the way they did. And it’s even easier to think that – because we have access to more and better information than our predecessors had – . Our forecasts will be more accurate than theirs.
But it doesn't matter how smart we are or how we carefully guard ourselves from our inherent bias and blind spots – we're human. And that means we're vulnerable to the same cognitive pitfalls as our professional ancestors.
Said simply, it’s easy to (incorrectly) assume that:
In retrospect, our own view of the potential future in 1996 likely would have amounted to predictions about how much data CDs and DVDs might store in the year 2022... “Hundreds of gigabytes, I bet!”
Instead, obviously, in a world where everything lives in the cloud and all media is streamed, the prospect of getting excited about local storage innovation is absurd; it’s irrelevant to the new paradigm.
So, the best we can do is attempt to minimize our biases, and prepare ourselves for a future where only one thing is certain: it won't be like what we're expecting.
Now, before we start making our own predictions, let’s briefly summarize what we actually mean by Artificial Intelligence...
As marketers, what we're really interested in is machine learning.
Machine learning is a subdiscipline within the broader field of AI "where systems have the ability to ‘learn’ through data, statistics, and trial and error in order to optimize processes and innovate at a quicker rate."
Simply put, machine learning is the process of empowering machines to learn from provided input, and then develop their intelligence and make perpetually better decisions.
We don't have to look very far to see examples of machine learning in action right now.
Consider the following:
To borrow a term from the world of Web3, those are just "Layer 1" implementations. Casting our net a little wider, we can see:
Bottom line: I think we can all agree that AI is already a game changer, and has far surpassed the capabilities envisioned by the smartest marketing minds just a couple of decades ago.
But we also know that it’s just in its infancy.
And so, the real question is... What comes next?
AI is an exponential technology.
And as humans, the fact is, we’re just bad at conceptualizing exponentials because we rarely encounter them.
It’s mind-blowing for us to see the results of things like compounding in financial outcomes. And biologically, we can see how the multiplication of bacteria follows this same pattern.
Exponential growth tends to surprise us because it shatters our linear expectations.
Taking that into account, let’s stretch our minds and try to imagine how AI might change our industry in profound ways in the near future.
Here’s what we see emerging in the next 6–12 months...
1. Endless Visuals
What: Utilizing next-gen AI to produce complex and detailed imagery based on verbal or written inputs.
How: While not widely accessible yet, tools like DALL-E, Deep Dream Generator, and Imagen, are starting to bring this capability to life, and will only continue becoming more powerful, and eventually, more mainstream.
Potential Applications: When access to these tools is democratized, the only limit to your visuals will be your imagination. Picture this:
Beyond this, what will it mean to be a graphic designer or animator in a world where images can be spoken (or typed) into reality – in a matter of moments?
And will designers go the way of the travel agent – or gain an exponential toolset?
2. Endless Text Generation
What: Leveraging GPT-3 powered applications to produce content and copy for any imaginable purpose, based on a handful of guiding inputs.
Potential Applications: Much like how tools such as DALL-E hint at the possibility of endless creative ideation, GPT-3 beckons similarly. Imagine this:
As with the impact DALL-E and similar tools will have on the graphic design industry… What will it mean to be a copywriter or content marketer in a world where AI can create a hundred different headlines – or even write a full essay – in just a few seconds?
3. Automatic, Pseudo-Native Translation
What: GPT-3 has the potential to be used for automatic translation. Over time, the fluency of this translation could approach near-native levels.
Potential Applications: How much would it be worth to your brand to be able to effortlessly communicate with potential customers in their native language?
While English is the dominant language in the Western world, just ~20% of the global population speaks it. And for three-quarters of those people, English is not their native language. Imagine the impact of being able to offer prospects a tailored experience, delivered in their natural dialect.
This will enable brands to legitimately scale globally, negating much of the prior language barrier in the process.
Again – each of these trends is already here. While their actual trajectory to become mainstream tech remains to be seen, there's no doubt that it will happen – nor their impact moving forward.
With these three foundational (and somewhat obvious) trends discussed, let's move on to some more speculative items.
Here’s what we see emerging in the next 2–4 years...
What: If GPT-3 can self-create content in written languages, what’s stopping a future AI from creating “content” (apps, task automations, etc.) in a programming language?
How: Tools like GitHub Copilot, YouCompleteMe, AutoCode, and Kite, while not yet capable of producing code at the level we’re describing, represent a step in the right direction, and a hint at what the future could hold.
In the future, we think it’s reasonable to assume that prompt-led software creation could become a reality.
“GitBot, build me an app that automatically hides all content about the Kardashians from all of my social feeds.”
(A guy can dream...)
Potential Applications: Imagine this: instead of laboriously documenting your technical requirements, hire a developer (or team of developers) to create what you need, and iterate on it for weeks - or months - until it works as required, what if you could issue a natural language command and have AI produce the first version of whatever you require in a fraction of the time?
And after you landed on a working, stable version, what if your sentient app could automatically maintain itself, keep up with external variables (API deprecations), update its various frameworks and coding languages, and in general just figure out how to stay functional on its own?
This is not as far-fetched as it seems… And the implications are enormous.
Being able to effortlessly create software for internal use, CX enhancement, or for widespread distribution would be as disruptive as anything we’ve talked about in this piece.
2. Self-Creating Campaigns
What: With powerful enough models and sufficient data to train the algorithms, AI could create and manage advertising campaigns from start to finish – no human intervention required.
How: We can already see how the combination of things – like DALL-E and GPT-3, along with the existing machine learning that already powers self-learning campaigns in Facebook and Google, could merge to bring this to life.
Potential Applications: The possibilities offered by this are limitless.
Imagine being able to create an infinite number of campaigns, each with hundreds of variables that can be tested, split-tested, and self-optimized over time… All while requiring nothing more than providing basic inputs (one time) and some simple, natural language guidance.
Moreover, what if these campaigns could be optimized down to the individual user level – creating a specific messaging sequence across multiple channels that are personalized to them?
This would have profound consequences – and it’s unclear whether it would tip things in favor of small players, or if we’ll just see advertising power further accrete to the largest brands as the cost of traffic goes parabolic.
We’re already seeing the giants move in this direction, with things like Google’s Performance Max service layer (which can already do things like auto-create video ads, etc.), but it’s still in its infancy.
3. AI Sales People, Customer Support Specialists, and More
What: As AI continues to advance, chatbots will grow increasingly sophisticated. With more and more of our interactions taking place in a virtual space, how long before we start to realize that the person on the other side of our chat window isn’t a person at all?
Potential Applications: Brands will be able to scale, serving more customers with less overhead. Human intervention will be saved only for those situations where it’s critical and most required.
This will enable low-touch, low-ticket products to be paired with more personal attention than was previously possible – at a limitless scale. Human resource issues (like employee availability, on-the-job training, and so on) will quickly become a thing of the past.
4. Personal AI “Butlers”
What: As AI assistants become available for personal use and increasing amounts of data are captured about each individual, personal AI butlers (in the vein of Iron Man’s iconic “JARVIS”) will be charged with stewarding their masters and caring for their well-being across all dimensions.
How: Natural language and open-source solutions like Mycroft are already available, and the range and quality of options will likely only continue to increase over time. These assistants will become trusted advisors, conducting research and aggregating data to help consumers come to complete buying decisions in record time.
Fitness trackers like Oura, WHOOP, and Fitbits will keep accumulating significant amounts of biometric information, giving users instant access to understand the implications of their behavior. Individual genetic data (tested with low-cost kits from companies like 24Genetics) will interface directly with these AI assistants. Wearable tech will allow intelligent algorithms to reliably predict your needs before you even know you have them.
Potential Applications: With ever-increasing amounts of data in the hands of your consumers, they’ll be in a strong position to make the right choice in any situation.
This could actually become an important sales channel, where the Butler entity could take on the role of an internal mentor, advisor, and concierge... Quite like a GPS, but for all aspects of your life.
Buying decisions and brand discovery might become largely influenced and driven by these Butlers, which surfaces a whole new marketing channel in the future – somehow getting your brand on the Butler’s radar.
Lastly, with all of this data on tap – much of it being sensitive –- it stands to reason that any good Butler would safeguard the privacy of their “master”, and act as a gatekeeper for them.
This could range from surfacing potentially unwanted data sharing to perpetually monitoring the web, looking for privacy breaches, or even acting as a vigilant data broker, once it’s possible for users to directly sell access to advertisers.
5. Auto-Generated Virtual Worlds
What: DALL-E, but taken to the next level. With nothing more than a series of natural language commands, AI could conceivably construct entire virtual worlds from the ground up. Images, text, textures, physics, underlying code – everything could be handled via a simple interface, refined as needed for any purpose.
How: Convergence of the tools already discussed here (visual assets, self-generated code, and so on) and further evolution of the tech involved would make this a very real possibility.
Potential Applications: While few people are sold on the idea that the world of “Ready Player One” will come to pass, there are countless applications for the creation of immersive digital experiences.
Just scratching the surface, a couple of examples include:
Bottom line: Having the power of world creation at your fingertips would be extraordinary.
We think most of the predictions above will come to fruition in some form or another - but they’ll almost certainly take shape in ways that we can’t envision currently.
This is the unpredictable, exponential nature of technological progress – which means that rather than betting on any specific technology, the best preparation for these coming changes is by being ready to adapt to the fundamental developments they will bring.
With that in mind, let’s talk about what you can actually do to plan for this oncoming wave of AI disruption (for better or worse)...
We've covered a lot of ground in this piece.
And hopefully, despite our inherent biases, we’ve managed to unearth at least a few ideas more than just extrapolating “today, but better.”
The fact is, AI has had a massive impact on the marketing industry already. And it's clearly going to keep having further impacts beyond what we can even conceive.
It’s worth emphasizing that Google and Facebook are in the fight of their lives to broaden the scope and accuracy of their AI modeling in the face of mounting challenges. Namely, the ceiling of observable conversion events continues to decrease thanks to things like iOS 14+, privacy regulations, and a cookieless future.
They have hundreds of billions to throw at this, and they will most likely solve these challenges and then some. Much of these advancements will be transferable to other technologies and use-cases... Ad infinitum.
So, with all the above in mind: where do you go from here? And how can you prepare for whatever lies ahead?
Here’s our take:
1. If you're not already collecting and storing your data in a usable format, start doing so immediately.
2. Familiar with the various consumer-grade AI tools available.
4. Understand those disruptive technologies are, by their very nature, non-linear.
5. Finally, understand that AI won't make a bad business good.
That AI will be the main character in the story of digital marketing over the next five years is practically a given.
The question is really where we’ll see the next breakout, exponential disruption. It’s impossible to say.
But you don’t need to know the future to be prepared for it.
So long as you ensure your organization has the core digital assets and the agility needed to pounce on these opportunities, you’re already well ahead of the pack.
And remember: if the campy sci-fi genre proves to be prophetic, once the AI starts taking over, it’s the Cyborgs who usually fare best. 😉
🗓 Chris Rempel is the serial entrepreneur and empty calendar enthusiast behind The Lazy Marketer
🚂 Eric Dyck is the cow catcher on the eCommerce express at DTC Newsletter and Podcast
P.S. Fun fact: every single image in this post was generated by DALL-E v2