Increased engagement. Personalisation ensures that consumers spend more time on vendors’ online sites because the more time they spend there, the higher the likelihood of purchases which in turn boosts sales. Vendor efficiency. Personalisation also eliminates the need for the vendor to manually try and guess what consumers might like. AI chat bots and efficient 24/7 support AI-powered chatbots operate round the clock, respond instantly and enable vendors to respond to multiple requests at the same time. Money is saved on hiring more people to respond to consumer queries. Furthermore, the really smart AI chatbots create a unique experience for a customer based on their past knowledge and interactions with consumers. Finally, if the data that chatbots rely on is accurate, a vendor can trust that a chatbot will tirelessly and efficiently churn out accurate information without the mood-swings or slip-ups of a human employee. Efficient advertising. Targeted advertising, averts wasting time and money on audiences that would not ordinarily purchase certain items. It boosts sales especially if they are received by people who are ready to buy. Furthermore, this tool is not only available to the big or medium-sized firms but to the small ones as well. This enables them to compete more favourably. These advantages should not blind us to the fact that there are serious downsides to the use of AI in E-commerce, as discussed below. The “Blackbox problem”. A customer utilizing the products or services of a business might encounter the “Blackbox problem”; a situation which arises from the fact that an AI model has produced an output which the customer cannot understand. Worse still, the vendor himself might not be able to explain why the AI’s decision that it made makes it hard to assess the correctness of a decision. The reason for this is that AI relies on intricate systems with numerous mathematical calculations, learns from patterns in voluminous amounts of data as opposed to simple instructions and then combines various chunks of information into one result. To illustrate this, imagine a small online clothing store which uses AI to tailor advertisements for its suede shoes on social media. The vendor has trained it on customer data comprising likes, follows, and purchases to target suede shoe lovers. Suddenly, it stops showing advertisements to consumer X who loves suede shoes and browses for them often, and instead targets Consumer Y who once liked a shoe advertisement. When the consumer X complains, the vendor checks but cannot explain why the AI has made that decision. Exploiting consumer weaknesses and behaviour prediction. If a vendor’s business is reliant on AI which knows a customer’s tastes and preferences, convincing consumers to purchase those products becomes easier. The customers are being persuaded to purchase something because AI is preying on their inherent tastes and preferences. Additionally, AI’s ability to predict enables vendors to exploit it to trigger customer behaviour. An example of this is choice architecture which simply means the way in which options are presented to people thereby influencing their decisions without actually restricting their ability to choose. These options are appealingly arranged to customers in a way that is likely to trigger the impulse to make purchases. Restricted offers and price discrimination. AI can also be used to restrict the offers that customers can have access to. This is done by presenting different offers to customers or different prices for the same product for different customers. By hiding some offers and pricing the same products differently, the idea of a free market is eroded. If this is discovered by consumers, it is likely to annoy, reduce sales and derail any loyalty that the vendor’s brand may have enjoyed. Biased decision making. As already highlighted, AI’s lifeblood is training data from which patterns are extracted to produce outputs in novel situations. If the data is biased and not representative of a wide range of people or if it has been selected because of its inherent biases, the AI’s decisions will appear biased. To illustrate this, assume a company that offers quick loans decides to use AI to decide who will benefit from its services. If the company trains its AI on records from the last ten years, mostly from consumers from urban areas with decently playing jobs and neglects to train it on information from people from rural areas, it is highly likely that the AI will determine the good borrowers as the ones who come from cities while those from rural areas will be perceived as risky. Apparent bias has the potential to ruin a vendor’s reputation and severely affect sales. Weakening self-determination. Consumerism is premised on the idea that there is a willing buyer and a willing seller. Ideally, the consumer has a range of choices on the market and the ability to make her own choice according to what she prefers. The consumer’s ability to make a choice relies on her access to knowledge but in reality, this is not always the case. As an example, European consumer law aims to protect this right by requiring that consumers are provided with sufficient information so that they can make intelligent choices. However, AI’s propensity to predict customer behaviour through vehicles like personalised offers and choice architecture affects the ability of a customer to remain in full control of her choices, thereby undermining the consumers’ ability to determine what is best for themselves. AI in commerce is groundbreaking, offering great benefits to both sellers and consumers in a way that was not thought possible many years ago, with the result that E-commerce is growing in leaps and bounds. However, the downsides cannot be ignored and the more that we are aware of them, the better we are at spotting them. Are we really in control or are we being controlled? By Nyombi Solomon Ignatius, (Associate Nyombi & Company Advocates) for Igikorwa Innovations Company Ltd.