Navigating the AI Revolution: Transforming Digital Advertising Beyond Buzzwords

Since ChatGPT’s AI revolution with its launch in late 2022 and massive success, tech world has entirely shifted its focus to Artificial Intelligence, and while almost entirely dismissing the previous hypes such as blockchain, crypto, web 3.0.

It is now clear that what OpenAI has been doing is groundbreaking in terms of giving ability to people access information at an incredible speed and quality unlike anything we have seen before. Even Google moved to ‘code red’ after the early success of ChatGPT and has been forced to make major progress on their version of AI bots¹.

Of course, this hype became another FOMO moment for the rest of the world. Every industry started using the famous words ML and AI in their offerings, because they, kind of, ‘had to’. Any company that did not have any ML/AI vision or mention those magical words in their earnings calls was seen as an ultimate fail by the investors.

As much as I agree that each company needs to truly incorporate the potential and power of ML/AI, most of them are just using these buzzwords in their short/long term plans for the sake of using them, instead of giving a clear explanation of how these technologies will change and improve their core offerings.

Digital advertising ,of course, did not lose a second, and all publishers, adtech companies, advertisers started using these buzzwords immediately. But, how can ML/AI realistically have the most impact in digital advertising? That is the purpose of this post.

First, lets have a clear definition of them both. Google² defines AI and ML as:

Artificial intelligence is a broad field, which refers to the use of technologies to build machines and computers that have the ability to mimic cognitive functions associated with human intelligence, such as being able to see, understand, and respond to spoken or written language, analyze data, make recommendations, and more.

Machine learning is a subset of artificial intelligence that automatically enables a machine or system to learn and improve from experience. Instead of explicit programming, machine learning uses algorithms to analyze large amounts of data, learn from the insights, and then make informed decisions.

Note: I will start using AI moving forward in this article as ML is a subset of AI.

As you can see from the definitions, AI is supposed to be able to mimic human intelligence by seeing, understanding, analyzing data, making recommendations, and responding to spoken or written language.

But, how? How can it do all of those these? With data, A LOT of training data to learn from, and making more and more precise predictions based on statistical models.

Basically, successful AI technologies like ChatGPT process huge amount of data and start making accurate predictions for the new prompts from users. They analyze, categorize, label lots and lots of images and texts, teach the machine how to interpret these inputs for what outputs, and eventually respond to any question by predicting the best answer to each question based on statistical models such as linear/multi-linear regression.

This means, not every automation or algorithm is AI, which is a common mistake and misunderstanding today. By telling a machine to do a certain task with a specific algorithm, it is not ‘predicting’ anything and ‘learning from the previous data. It simply completes a task. For example, if you tell a machine to increase bids for an ad or budget when the ROAS is above a certain threshold, it is not AI, the machine is basically following a pre-defined rule. However, if you tell the machine to learn all the patterns of different campaigns, creatives, and eventually come up with a new and optimized budget allocation, creatives, targeting options etc, this is where AI shows its true potential.

So, lets talk about what type of things AI can do in the new era of digital advertising:

AI Revolution in Measurement

Measurement will probably be the most obvious and powerful utilization of AI in the near future. Especially since cookies and other identifiers are going away³ (already dysfunctional in iOS, and will be so for Chrome and Android soon), which were historically used for people-based attribution models, AI can make a huge impact by analyzing the massive amounts of historical campaign data (impression, reach, click, conversion, lifetime value etc) from different channels and providing recommendations of best media allocation and campaign optimization for maximum impact.

For example, open source MMM models already started utilizing AI to provide the best recommendations and visibility on campaign performance measurement to advertisers.

From Meta’s open source Robyn⁴:

The Meta Marketing Science team has built Project Robyn, an open-source R code that uses machine learning techniques for in-house and DIY modelers that clients can use to build in-house models.

Without accurate measurement, none of the other components of digital advertising can not be utilized at maximum capacity. Therefore, more accurate measurement with AI will be the pinnacle of its impact on digital advertising.

AI Revolution in Advertising Creatives

With Generative AI, which means the use of AI to create completely new content, like text, images, music, audio, and videos, advertising creative ecosystem is entering a whole new era.

Creative production has always been a costly process with relatively long delivery cycles for advertisers. This prevented effective optimization as advertisers could create only a limited number of different high quality versions of their creatives. Even for e-commerce advertisers, which utilized some dynamic automation tools creative optimization purposes, their ability was also limited with the images of the products and filtering/framing.

Generative AI will make a revolution in advertising creatives in the near future. With more realistic, and higher quality videos, images without needing to produce net new content, advertisers will be able to test and optimize the best creative for their audience dynamically. They will be able to run different creatives for the same message for different audiences to optimize their campaigns real-time.

For example, Coca Cola ran ‘Create Real Magic’ campaign last year which was entirely created by OpenAI’s Generative AI technology, called Dall-E. This campaign was more of a contest, then a direct advertising campaign but it will not take too long for them to start utiilizing AI for creatives.

For more images: createrealmagic.com

Recently, OpenAI announced a new service called Sora⁵, which creates AI-generated videos based on text-based prompts. It is not hard to see how it will eventually change the digital advertising space as we know it.

AI Revolution in Targeting

Targeting is also heavily affected by the iOS14 changes and soon-to-be fully in effect Google’s deprecation of cookies in Chrome. These changes mean advertisers can not target people across different websites or apps real-time without explicit consent from users, and will have to rely more on their CRM data (aka PII – personally identifiable information). There are companies such as Liveramp which does identity matching, however AI will provide tremendous value here as well.

At the end of the day, effective targeting is all about clustering people in the groups with similar attributes (e.g intent, demographics, income levels etc). AI has proven to be extremely effective at finding patterns in data and cluster groups for classification. This means advertisers will be able to utilize AI-powered targeting solutions to target the right ‘clusters’ without having to need direct access to all or most PII or cookies, which will also contribute to a more privacy-friendly digital advertising ecosystem.

It is important to note that such effectiveness can only be achieved with huge amounts of data as well as the metrics for the business outcomes, such as conversion, or new customer. Thus, data aggregators and big tech will have significant advantage over other players or individual advertisers in this space.

AI Revolution in Campaign / Media Optimization

Optimizing campaign features and media allocation is one of the fundamentals of digital marketing, as we all know. Until now, these operations were manual or semi-automatic through API tools through defining pre-defined rules. However, these semi-automation capabilities were not truly ‘learning’ on their own and making decisions/changes in the campaign itself. Now, with AI, campaigns will be run by machine learning algorithms which will test and learn the best possible combinations of campaign budget, bid, optimization type, creative (potentially created by Generative AI), and target audience.

Similarly, digital marketers have been manually monitoring campaign performance across different channels and re-allocate their budget on a regular basis across best performing ones. AI will be able to automate this process as well beyond campaign optimization and execute on optimizing budget allocation. AI will ensure that the campaigns are not only setup fully optimized for best performance, but also overall budget is spent optimally across all publishers.

Today, especially for campaign optimization, big tech publishers such as Google and Meta made huge improvements (e.g Google Performance Max, Meta Advantage). Thanks to their vast amount of historical data, as well as real-time people or intent data, they were able to decrease their dependency on third party data which was heavily affected by Apple iOS and Chrome/Android changes to cookies/proxies.

Customer Support and Social Media Monitoring

Chatbots have been around for a long time, however they rarely did what they are supposed or expected to do: support customers effectively for their issues like a human. They were merely able to provide FAQ articles from help center and direct the customers to an agent.

ChatGPT and other AI-powered chatbots will absolutely change that. They will ‘imitate’ human support to customers and provide detailed explanations and guidance on issues or provide recommendations. Instead of scrolling through pages of products, customers will be able to basically tell the AI bot what they are looking for, and they will access the information instantly. The bot will not only provide what the customer wants, but also provide recommendations more effectively, which will help customers discover new products, and increase basket-size for the retailers. Today, these bots are utilized on websites and apps, but in the future, they will be an integral part of shopping experiences in the stores as well.

When it comes to social media moderations, there were auto-reply bots already, trying to imitate humans but they were simply copy/paste pre-defined responses, which people already recognize and ignore, without truly processing the messages/tweets and customizing responses accordingly. However, in the future, with the deeper integration of AI bots, they will be able to provide custom responses to the people in the social media comments or direct messages, and engage with them discussions on problems, questions and issues. This will make the social media management much more automated, effective and personalized.

Risks & Mitigation of AI in Digital Advertising

All of these and more implications of AI are exciting developments for digital marketing, indeed. Every aspect of digital marketing will be impacted with the magical touch of AI. However, this will not happen without pains. Here are the risks I am foreseeing:

Human vs machine interaction as part of customer & company relationship: As seamless as AI bots can and will be, I think there will always be a factor of errors with the machines which will create doubts in effectiveness. At the moment, there are active discussions around ethics, bias, privacy and copyrights, which put AI applications under the spotlight. Such problems can be tolerated for a cutting edge company like OpenAI, but for businesses, where their brand image is at stake, they will probably be much more cautious. There are already examples of AI bot failures reported since early adoption of AI chatbots by companies.
- In order to mitigate this, my recommendation is for retailers to think about their risk tolerance level and categorize these applications of AI accordingly before adopting them. For example, utilization of AI in ads measurement can be categorized as zero/low risk for the retailers, but in creatives, it should be seen medium/high risk. This does not mean they should not utilize AI for creatives, but they should be more cautious in terms of automation and moderation of which creatives will be live and optimized real time.
Need for mass data to effectively optimize AI: As mentioned earlier in the post, AI models need huge amounts of data to be able to effectively perform tasks. While OpenAI is capturing data from all over the internet by crawling or accessing public data to train their models, each retailer / company will not have such data for their own custom models to perform as effectively.
- Companies will probably need to train their custom AI models differently with more supervision to decrease the errors. Also, I think there will be aggregators which will consolidate data across different companies within the same industry to train AI models more effectively and benefit all companies within that industry at the same time.
Capturing more, potentially personal data: AI bots already use any data captured through the user prompts to optimize their models.
- From OpenAI help center: “We may use content submitted to ChatGPT, DALL. E, and our other services for individuals to improve model performance. For example, depending on a user’s settings, we may use the user’s prompts, the model’s responses, and other content such as images and files to improve model performance.”
- This means retailers will capture more data about their customers than ever before and it will get increasingly more important how they will store and use that data. Think for a second that all your discussions with the sales person in the store is saved by the company. Yes, similar data such as browsing behavior or call center calls were saved in the past, but this data volume will potentially be much larger, and likely more personal, given the interactions can go deeper with these AI bots.
- Retailers should ensure that before they implement these features, they have a clear plan on how to store this data, and disclose to their customers how they will use it to avoid any issues in the future. This happened in the cookie era: at first, companies collected all the data with little to no disclosure of how they would use it, but regulatory bodies eventually went after them and enforced disclosures and consent flows to avoid misuse of data by the companies.

Closing Thoughts

I have been following digital ecosystem since early 2000s, and there has been a lot of revolutionary changes since then: Web 2.0, mobile, social media, blockchain and now AI. Among all, I think AI will make the most impact to larger group of people and companies, and increase both effectiveness and efficiency in particularly digital advertising, but will definitely introduce new problems. I am hoping that as an industry, we will utilize the benefits of this new revolution while avoiding the shortcomings as much as possible.

References:

Disclaimer: title of this post was created by ChatGPT