Aurora Magazine

Promoting excellence in advertising

The Good, the Bad and the Ugly

AI is here to stay, whichever way you look at it, argues Syed Amir Haleem.
Published 04 Jul, 2023 01:11pm

Dr Geoffrey Hinton, the man considered to be the godfather of AI, recently quit his job at Google in order to voice his concerns about the technology. Hinton is a deep learning pioneer who contributed to the development of some of the key techniques at the core of AI. He helped set up Google Brain, one of Google’s flagship AI projects. And Hinton is not the only one. Elon Musk, Sting, and the US government have also raised red flags about AI recently.

However, the question we are concerned with here is the impact of AI on marketing, and it would be foolish to believe that just because we live in Pakistan, the technology will impact us later than the rest of the advanced world. Slow adoption of technology is a thing of the past; today, new technology is available to everyone, no matter where. In my opinion, there are three ways that AI will impact marketing – the good way, the bad way, and the ugly way.

1. The Good Way

l Segmentation/Targeting: Customer segmentation and targeting are two major areas where AI has the potential to make a difference right now. AI enables marketers to segment their target audiences, analyse customer data and personalise campaigns. The result will be better engagement, higher conversion rates and greater customer loyalty. When digital first came to Pakistan, brands could focus on specific audiences; for example, young male adults living in Clifton in Karachi with very specific demographics and selected digital media consumption habits. This was a huge deal considering the alternative of placing an ad on network TV and hoping it would grab some of the target audience’s eyeballs.

l A/B Testing: AI systems have taken this a step further by building knowledge databases that keep track of how each individual in an audience cluster responds to a type of product or a type of colour. Hypothetically, AI can tell us that 67% of that audience in Clifton responds better if the colour red is used. Furthermore, AI automates marketing processes like content production, social media management and email campaigns, thereby saving marketers time and money, allowing them to concentrate on higher-value initiatives. Even today, we have access to AI systems that can tell us if certain words get a better conversion rate in an ad copy – and with more learning, an AI programme will be able to write ad copy that delivers better conversion.

l Customer Relationship Management (CRM) Chatbots and virtual assistants can offer personalised customer assistance, enhancing customer satisfaction and reducing labour costs. This is not a long-term prediction; as we speak, there is probably an ‘aunty’ somewhere in Defence chatting to an AI-powered chatbot and ordering her next lawn outfit. Seasoned chatbots have been known to reduce customer service costs by up to 30% – this could translate to over $23 billion saved in the US market alone. Additionally, chatbots don’t need sleep, meaning that you have a customer support system that works 24/7, which is very useful if you plan to serve markets in different time zones.

l Content: The most talked about impact of AI is its ability to create content out of thin air. It can create an instantly likeable brand ambassador that will not charge absurd amounts of money, throw celebrity tantrums, refuse to say what needs to be said or be late for shoots! Unless there is legislation stopping people from using this technology, we will soon have AI actors who become huge celebrities but aren’t actually real.

2. The Bad

l Quirky Technology: Last week I registered a complaint with a local smartwatch e-commerce website. A chatbot popped up and the conversation went like this:

A female AI assistant introduced herself as Mariam and asked me how she could be of assistance.

Me: “I have issues with my watch.” (Not yet realising this was a chatbot.)

Mariam: “What would you like to watch?”

Me: “I would like to receive messages from my smartphone on my watch, but the function won’t work.”

Mariam: “You can’t watch that.”

Me: “Excuse me? It said on the box that I can.”

Mariam: “You are excused.”

Me: (realising now that this is AI) “I would like to make a complaint about my watch.” (Starting over in simple terms.)

Mariam: “Okay. What would you like to watch?”

I gave up, deciding it would be easier to visit the shop in Khadda Market rather than get through to Mariam the chatbot. Yes, the technology may not be perfect, but it will get there fast, because these systems start off simple, but learn and improve with each conversation. The more interaction they have, the more effective they become.

l The Data Dilemma: AI systems are based on two separate parts – the code and the data. While the code is the same for everyone and easily available to anyone globally, AI systems give the best results when they are loaded with data. Let me explain how this works. Take a successful A/B testing AI system – it is a simple programme that displays two different creative executions for the same campaign. It then tries to learn which creative gets a better response in order to teach itself. Now, even if this tool has a track record of creating ad copy that delivers great conversion rates in the US, it might not perform that well in Pakistan. This is because the system has based all its calculations on data derived from an American-English speaking market with a different culture and way of thinking. Making matters more complex is the fact that the Pakistani market also has Minglish (a mix of Urdu and English), and on top of that, its own idiosyncrasies. For example, the word ‘typical’ is used colloquially to describe something that is ‘atypical’. It is a common error, but highly prevalent.

l AI Skills: Most marketers in Pakistan, when they think of AI, think of an intern writing content using ChatGPT. In reality, AI systems are extremely complex and require a specific skillset. There has been progress in media planning because of the investments made by MNCs. However, areas such as AI based content creation have not been so blessed by this kind of investment.

l Speech Patterns: Although AI speech patterns have evolved, they still have an artificial, robotic quality to them. This is a worry shared by the digital marketing community.

l Similar Content: The responses that ChatGPT generates can appear remarkable when you first try it out. They are interesting, diverse, and sometimes even human-like. But if you keep repeating those prompts, the material becomes monotonous and even boring because the same information is repeated, which means that AI-based content could end up sounding similar to other AI-produced content.

3. The Ugly

l Deepfakes: Once the technology is perfected, you can make a video of anyone doing anything. Dictators could use the technology to create negative content about their opponents. In fact, anyone with a vendetta could damage someone’s reputation by creating deepfake videos. In the last US election campaign, videos of Joe Biden were circulated that said things Biden never said. Thankfully, they were so badly produced that they were easily dismissed.

l Job Loss: Four years ago I was transiting in the Middle East and met a French software engineer in charge of automating a plant in Saudi Arabia, who said he had reduced his headcount by 15%. I was immediately interested and struck up a conversation about the impact of AI on the workforce. The gist of our conversation was that the impact will be huge and it will happen faster than anybody thinks possible. I agreed because I have been working with AI and I know the potential damage it can cause. Some of the people who had joined our conversation thought this was still science fiction. Well, five days ago the BBC ran an article about layoffs by British Telecom. The company has started the process of reducing its workforce by 55,000 employees by the end of the decade. A fifth of those jobs (11,000 employees) would be replaced by AI. So, 11,000 people in one company alone are about to lose their job in the next one to seven years and there is no way to stop or reverse this process. Imagine the impact on the job market when every company is on board with AI. Imagine a fifth of the global workforce unemployed in the next seven years – and companies already have the technology to do so.

To sum up, AI will solve many problems but it will also create many new problems – for which we have no solutions right now. As marketers we will have many opportunities to use AI to grow and scale businesses but I also fear that in many instances we will be fighting AI as well. The need of the hour is for everyone to become acquainted with what AI can do for us – the good, the bad and the ugly.

Syed Amir Haleem is CEO Skale Interactive and Content Fission.