The most important part of AI is the human intelligence behind it
Artificial intelligence is a powerful and exciting technology with the potential to revolutionise many industries, including marketing. However, human intelligence ultimately determines how effectively AI is used to drive real-life benefits.
![](https://marketingweek.imgix.net/content/uploads/2023/01/26111434/Shutterstock_1502881307.jpg?auto=compress,enhance,format,redeye&crop=faces,entropy,edges&fit=crop&w=736&h=429)
Technology, data, machine learning and artificial intelligence (AI) have all played a significant role in shaping the marketing industry. They continue to move forward, enabling brands to gain insights into customer behaviour and preferences, personalise marketing campaigns, improve customer engagement, predict customer behaviour and more. But without human intelligence overseeing these technologies, models and algorithms are at best useless. And at worst, dangerous.
The importance of this can’t be overestimated. The goal of AI is to create machines that mimic human intelligence, but without necessarily relying on it or making decisions in the same way. Human thinking must remain at the heart of AI to ensure that systems align with human values; that they are transparent, accountable and beneficial; and that any negative consequences are identified and mitigated.
If you’re looking to use AI in future marketing efforts, it is essential to ensure that human intelligence leads the thinking in the following areas:
1. Know what you want to know
AI can’t understand what you need to know. If the question is not well defined and founded in a deep understanding of the specific problem or goal that you want to address, the answer will not drive the required decision and action.
To quote Einstein: “If I had an hour to solve a problem and my life depended on the solution, I would spend the first 55 minutes determining the proper question to ask.”
The first step in driving marketing effectiveness is to know what you want to know. What is the challenge to be solved, and what information will enable constructive changes to be made?
The starting point for any implementation must then be to understand what data is needed to train and test the AI model, enabling it to reveal actionable insights that inform decision making. There is no value in having lots of data points if you don’t know what to do with them or how they are relevant. Ensure that your quest to gather the data required focuses on revealing actionable insights.
2. Design AI with insight so that it’s fit for purpose
Designing an AI solution that is fit for purpose is crucial to making positive business decisions. The process of building models and technologies is both an art and a science and requires expert understanding of the problem or goal that you want to address.
Marketers have often asked me why previously used econometric models didn’t provide the accuracy or predictability they expected. The answer is the human modeller.
When designing an AI solution, it is important to have a team of experts who understand the world they are modelling and have a depth of understanding in the specific application. At Jaywing, data scientists collaborate in teams alongside marketers, leading to unique model designs. For example, the inclusion of base sales in an attribution model ensures that the true incrementality of marketing is measured. This overcomes a common industry challenge, where the credit allocated to marketing activity is otherwise inflated, thus relieving marketers of awkward discussions with finance.
The process of building models and technologies is both an art and a science, and requires expert understanding of the problem or goal.
It is also important to understand the context in which the AI model is being built. Different techniques can be used depending on the problem or goal at hand. For example, controlled A/B experiments are considered the gold standard for measuring impact, but they can only test one or two things at a time. In the world of marketing, there are often multiple activities that need to be understood.
In this case, employing Random Forest methodology, which harnesses the natural impact of tests and variance to measure the impact of every variant across thousands of journeys, has proved to be a better solution. This domain understanding helps ensure an AI solution is fit for purpose and can deliver the insights needed to make positive decisions.
3. Interpreting outputs
When mathematician Abraham Wald studied data of aircraft returning from World War II to decide where best to place limited extra armour, his interpretation of what he saw turned previous thinking on its head. The data showed that for returning planes, the places most damaged and therefore perceived as requiring additional armour were the wings, nose and tail. Wald offered an alternative analysis – that the military didn’t need to reinforce the spots that had bullet holes. They needed to reinforce the spots that didn’t have bullet holes, because if the planes were shot in these bullet-free zones, they didn’t return home.
This is an example of when the results reveal the opposite to the obvious conclusion. It is easy to make bad inferences with data, which is why the expertise of humans who have the benefit of context and critical thinking is vital to ensure that the work of AI technology is harnessed correctly. Marketing is about people and psychology – no single metric, no matter how accurate, can provide the answer to how successful activity has been, or reveal what to do next. Marketers must work in partnership with analysts to draw answers together from a suite of techniques employed to build a fuller picture of effectiveness.
Humans can understand the data and context that models don’t. Analysts with years of experience will grasp predictability and accuracy and eliminate potential AI bias. This is vital for marketers planning to take actions with AI. Take for example the AI model Amazon designed to select the best resumes during its hiring process, but which subsequently showed a preference for male candidates. It was human intelligence that identified this significant failing and prevented biased hiring.
4. Looking to the future
It’s important to note that AI refers to a system developed to perform a specific task. In the case of ChatGPT, for example, this is to respond to a natural language request with a coherent and well written answer based on the information it was trained on. This should not be confused with artificial general intelligence (AGI), which can solve any problem in the same way a human can, and which most experts believe is still many years or even decades away.
AI may mimic human intelligence, but the definition of the problem it solves, the choice of training data and the use cases all rely on humans. AI can interrogate and produce insights from huge datasets well beyond any human capability, but it lacks the context and understanding that comes from a human experience of the world. So human thinking must remain at the heart of AI for now.
AI is a powerful tool that will continue to revolutionise the marketing industry, but its effectiveness relies on human intelligence. Insights provided by AI are only valuable if they are not only fuelled by data and understanding, but also used to evolve marketing plans and improve communications, as well as to drive positive change for both customers and the business. Without human action, AI is just a collection of data and algorithms – it needs human intelligence to turn insight into action and make real impact.
Dr Catherine Kelly is managing director of media and science at Jaywing.
Need a different kind of agency? Let’s talk.