The role of AI in marketing is anything but new. Various companies, ranging from Google, Facebook, TikTok, Microsoft (the bigger ones we hear of every day), and several other providers of various advertising technologies and content functions, have used AI extensively over the past 5+ years. The capabilities have only gotten better with time as processing, storage, and network access have constantly become cheaper, with access to device-level data and the ability to stitch together several sources of unstructured data.

With the public appearance of tools like Chat GPT, a new wave of generative AI has unveiled itself to the public. However, what is important is that: this is not new. Several companies have been working on the text, and transcription, including an AI-based news writing company in Pakistan, as well. But this has rarely been available to the public directly, with a vast information source connected to it in a conversational style.

Chatbots have already existed with varying degrees of language capabilities and information source depth for a while, and some AI-based content generators have taken the lead, making them popular with ChatGPT.

The advent of cheap, abundant AI-based content platforms opens to leave the local marketing landscape with interesting challenges. Firstly, in a market severely short of good copywriters (for short-form advertising copy), will the mid-tier and smaller agencies be more inclined to lean on AI-based tools for copy generation? I would typically expect the large ones to retain a slightly more purist approach to copy, but anything is possible.

The second question is, who owns the copyright to such content that has been AI-generated?

The governance structure around AI-based content is a tricky space globally as well. Just for context, a company that created a robotic lawyer that uses AI to help defendants fight traffic tickets cancelled its plans after actual law firms sued it because the robot lawyer doesn’t have a degree or a license, which, of course, is part of the requirements to be able to argue in court.

There are fundamentally different types of AI formats at play here.

The first (and lately, more popular) Generative AI is centred around creating text, images, videos, and sounds based on various learning models.

The second, which is slightly hidden from the public eye but is really much more spread out across the business space, is non-generative or standard AI which is used primarily across learning models, predictions, and analytics. Why do I say this is more pervasive? Because it is embedded in every phone these days! Both Apple and Google have AI engines built straight into their phones to handle a chunk of predictive use of the devices. Siri, Alex, and Google Assistant are strongly AI-powered tools that we interact with on a daily basis, aside from their ad-tech-related tools.

Other examples include shopping recommendation engines, dress size mappers, customer support tools, call routing engines, refrigerators that self-stock and suggest healthy recipes, and, how can we forget, self-driving cars!

I digress. Back to AI, Marketing, and Pakistan. When it comes to Standard / Non-Generative AI, we already have some implementations in place by virtue of tools we buy and use from players around the world.

Pakistan may not really feel the real force of AI in marketing until our market players are integrated through and through. Of course, the uses for generational AI, which rely on external stimuli and information parameters, will continue to work; however, in the absence of detailed personal profiles, the expectation for content to be specifically personalized remains limited.

In developed economies, data brokers hold huge sway in the direct-to-consumer marketing segment through collection and data enrichment offerings that connect through banks, credit bureaus, utilities, and municipalities. The core data component for profile creation is mostly through owned consumer data, i.e. 1st, party or zero-party data, of which Pakistan faces a severe lack. Arguably one can say that all e-commerce players have 1st party data on their consumers, but that’s only across their own systems with no enrichment mechanisms from 3rd party sources.

The absence of enrichment sources or the absence of deeper information about these consumers leads to limitations in learning about customers, which helps artificial intelligence create content/offerings for customers.

One key example is an extension of the ad networks algorithms where a certain creative is shown to a user out of the few uploaded that are expected to drive the conversion. Think of this on a newer level, where both creative text, copy, and landing pages are customized based on the potential target consumer. Landing pages are already being customized. The ads, not yet.

Customer support and product cross-sell, up-sell, and replenishment cycles can also be handled better. However, these are true scenarios for D2C brands and platforms.

With CPG / FMCG swiftly adopting online sales as well, the avenues open up to drive higher personalization in ads and LTV management through optimizing top-of-mind recall closer to the repurchase event in the consumers’ life.

This means that a user may see an ad for a product on YouTube because they are a consumer and they are expected to purchase soon, or the brand may try to up-sell them a different product or target them based on their stage in life. (eg. infant formula and diapers initially, followed by toddler products and growth-related ads later in life).

There is, and will always be, contextual information on customers about transactional history and information about their browsing behaviour, and some visibility on their social profiles which will help create highly personalized communication.

Unfortunately, aside from content level optimizations (or shortcuts by some agencies to use these tools instead of humans with a deeper product insight), the data inputs to refine content further, instead of just sales data, seem to be a while away for Pakistan. This is owing to two primary reasons.

  1. The data network infrastructure is data intensive, and the new world depends largely depends on it. The tragic part is that it is highly concentrated in the urban areas. Step outside of a tier 1 city, and you can visibly see network performance drop.
  2. The data feedback loops needed to serve as inputs to these algorithms are missing. There are some data points, as mentioned above, but by and large, there will have to be a great deal of approximation.

So the machines are here, they’re here to stay, for good, but they won’t be rising up anytime soon. There will be a few blips here and there, but before any of that, there will definitely be a lot of agencies and brands who will want to jazz up their presentation by adding AI as a term.

Previous articleCopy, Paste, and Beyond: A Creative Adventure with Generative AI
Next articleWhen Communication Collides with Technology