Marketing Now A Part Of AI Governing
A committee, team or body is responsible for overseeing the development, deployment and use of artificial intelligence programs. One must be created if you don’t already have one.
In my previous article, I discussed the main areas of AI and ML modeling in marketing. How these models can help you innovate and meet client needs . This article focuses on marketing’s role in AI governance.
What is AI Governance?
AI Governance is the process or framework that governs your use and adoption of AI. Any AI governance effort should aim to reduce the risks associated with AI. Organizations must have a process to assess the ethical implications of AI-driven algorithms.
Industry is a major factor in determining the level of governance. A deployment of AI algorithms in financial settings could pose greater risks than one that is used in manufacturing. An AI algorithm that allocates parts economically around a plant floor requires more transparency and oversight when using AI to assign consumer credit scores.
An AI Governance Program should consider three aspects of AI-driven apps in order to effectively manage risk.
- AI Governance Data — What data does the algorithm use? Is the quality of the model appropriate? Are data scientists able to access the data required? Privacy will be violated by the algorithm? (Though this is not an intentional act, AI models may accidentally expose sensitive information. Data can change over time so it is important to maintain consistency in data use within the AI/ML model.
- AI Governance Algorithms — Does the data change affect the output of an algorithm? If a model is created to predict the customers’ purchases in the coming month, then the data will begin to age and impact the model’s output. Are the appropriate actions or responses still being generated by the model? Marketers need to be aware of model drift, as machine learning is the most popular AI model in marketing. Any change in the predictions of the model is called model drift. The model is considered to have “drifted” if it predicts something different today than what it predicted yesterday.
- AI Governance Use. Are the people using the AI model’s outputs been trained in how to use them? Are they evaluating outputs for errors or if there are any deviations? This is particularly important if an AI model generates actions that marketing uses. The model should also identify the most likely customers to buy in the following month. Have you taught sales and support representatives how to deal with customers most likely to purchase? Is your website “aware” of what to do with these customers? This information can impact your marketing processes.
What structure should it take and who should participate?
AI governance can be organized in many ways. These approaches range from highly controlled to autonomously monitored, and it is highly dependent upon the industry and the corporate culture.
Governance teams are usually made up of technical staff who know how algorithms work and leaders who can explain why models should function as planned. The governance structure usually includes someone who represents the internal audit function.
Whatever the structure of AI governance, the primary goal should be collaboration to ensure that AI algorithms and data are used in a way that is compliant with all regulations.
This is an example of AI Governance, which can be used in highly regulated industries such as healthcare, fintech and telecommunications.
What can marketers do to improve AI governance?
Marketing can be involved in governance of AI models for many reasons. These reasons all relate to marketing’s mission.
- Advocate for customers. Marketing is responsible for ensuring customers have all the information they need to make purchases and to continue buying. Marketing is responsible for customers’ experience and protecting customers’ information. These responsibilities require that the marketing department be involved with any AI algorithm that uses customer data or any algorithm that has an effect on customer satisfaction, purchasing behavior, advocacy, or advocacy.
- Brand protection Protecting the brand is a primary responsibility of marketing. Marketing should intervene if AI models are used in a way that could damage the brand’s image. Marketing should play a key role in the deployment of AI models. For instance, AI-generated creditworthiness scores can be used to decide in advance which customers will get the “family discount”. Marketing should be part the team that determines whether the model will produce the desired results. Marketing should always ask “Will this change the perception of our primary customers about doing business with me?”
- Open communication. Storytelling is an essential part of AI/ML model creation and deployment. It is often overlooked. This is necessary to make others understand the purpose of the models. Good, well-governed AI/ML modeling must be transparent and easily understood. Transparency is the ability to explain the models created and used by the creators and the users, as well as the leaders and managers of the organizations. The AI Governance team is at great risk of not being able explain what the model does to business leaders and how it does so to stockholders, government regulators, counsels, or outside counsels. Marketing is responsible for communicating the “story” about the model and its implications to the business.
- Guarding marketing-deployed AI Models. Marketing should be a major user of AI/ML models that help to determine which customers will buy the most, which customers will stay customers for the longest, and which customers are most likely to recommend your company to other potential customers. Marketing should be able to sit at the AI Governance table in order to ensure that customer information and privacy are protected.
First, learn the basics
Although I’m confident that your AI Governance team will welcome marketers to their table, it is a good idea to be prepared. Before you get started, here are some skills and capabilities that you should be familiar with:
- AI/ML understanding. Understanding AI/ML and their workings is essential. While this does not imply that you should have a PhD in data science, it is worth taking an online course to learn more about these capabilities and their uses. You need to understand the impact of the models, especially if it exposes customer information or puts the company at risk.
- Data. The first place that bias can enter an AI model is in selecting and curating data. If you want to analyze customer behavior regarding a product, for example, you will need three-quarters the data. This data must be collected in the same manner and curated so you have accurate as well as complete information. Your role in marketing data is even more critical.
- Process. The process by which the algorithm is to be implemented should be well understood. As a marketing representative, you need to be familiar with the process of how the algorithm will be deployed. Many marketing teams will appoint their head of marketing operations because this is an essential skill.
It doesn’t matter what role you have in AI Governance. It is imperative that AI/ML be used responsibly within your organization. This requires persistence and vigilance as the models learn from the data.