While the term AI, artificial intelligence, has stirred a great deal of excitement, it is not always worth the hype. People tend to confuse two completely different concepts: narrow and broad AI. As the name suggests, narrow AI is very limited. It can perform a single task, and though it cannot do that task perfectly, it can sometimes outperform a human being. The term broad or general AI, on the other hand, refers to human-like intelligence. Broad AI is a thought-provoking concept that makes for good movies, but for the foreseeable future, not much else.
To make better decisions, we need data. To get even more out of the data, we need artificial intelligence. The value of data is something no company should ignore. With AI, I mean it in its narrow meaning: using data analysis and machine learning. Supporting decision-making and task automation is where AI can already outdo us in terms of speed and accuracy.
It’s good to remember that AI is mainly used to do the exact same things that marketers are already familiar with. AI can be used for targeted marketing, budget allocations, or creative content design, for example.
The benefits of AI should not be exaggerated, but they should not be understated either. Many things that could not be done accurately enough before, such as multilingual machine translation or image recognition, can now be achieved with the help of AI.
If you are new to AI in marketing, you should start with short trials. A successful first trial might not be the most profitable one but the one that rapidly produces tangible results and is not too complicated to implement. The projects should always be part of a larger whole: don’t experiment just for its own sake. And don’t start too ambitiously. For example, if you start with a project to store all the available data, you may get bogged down by details and lose focus. Instead, ask yourself what you could achieve now. The first projects will teach you about the nature of AI projects, give you momentum, and make your organization better equipped for larger-scale AI initiatives.
AI projects can be roughly divided into two categories. First, there are one-time projects that deliver insights and can be summarized in a few slides. One-time projects can be of great value if they generate additional information to tackle an issue or create future opportunities.
Then there are software projects that result in a system that constantly runs in the background carrying out analyses, decisions, or optimization. If you’re either having a hard time choosing what to do with data or drowning in options, consider the decisions you need to do next. These kinds of decision points are logical candidates for data analysis projects.
To ensure that the project is useful, the insights from data should lead to actions, or the software system should be put into production. With one-time projects, it is essential to think about what questions you want the data to answer and what you will do with that information. It happens all too often that successful proof of concepts never goes into production, or that data insights do not lead to actions. The result is a never-ending series of proof of concepts.
Traditional organizations usually like to have their solution delivered in a few months with great certainty and according to the quality requirements. It’s good to remember that with AI and data, the desired performance is not an absolute certainty. If the amount of data is insufficient or its quality is lacking, the project’s desired outcomes might not be achievable. Data analysis might still give you other insights or new ideas about how to use the data.
AI projects are not traditional projects. If the organization does not have previous experience of agile methods, the nature of the development project can come as a surprise. The first step is to analyze the data. At this point, some people will get anxious: “Why are we spending days or weeks analyzing data, shouldn’t we get started on doing things?” After the analysis, the first baseline solution will prove that the data has the predictive power needed. From there, you can advance sprint by sprint, improving the solution as long as the results are worth the effort.
In the best-case scenario, a positive circle is created. The AI solution brings more customers who provide more data, which can make the solution even better.
The basic principles of AI are not that complicated. As AI is already an everyday thing in marketing, every marketer should understand those principles. Coursera’s AI for Everyone is a free online course that’s great for learning how to use AI.
*This blog post was originally published in collaboration with ASML.