Companies often extoll the benefits of the latest artificial intelligence (AI) systems that they have put in place to help accelerate a business process, making it easier for workers and greatly enhancing productivity for the company. However, all too often, companies assume that they can take any product anywhere and it will be as successful. It is important to remember that not all countries, people and processes are the same.
Take sales and marketing AI systems; in the U.S. and UK, buyers often display similar signals and behaviours that can identify them as looking to purchase a product, but these can be completely different for a person in Germany or Japan or Brazil.
This means that if the same AI is used in these different locations, it might give inaccurate recommendations and predictions compared to its original home market, making it harder for teams to achieve their targets rather than easier. It will also have a negative impact on the company’s growth as leads are missed or important decisional data points are ignored.
Variety is crucial to AI
Over the last decade, the U.S. has produced the most AI journal publications, closely followed by China. It has also led the world on business and academic collaborations on AI papers, more than double that of the EU and China. This means that the majority of AI systems for businesses are based on work done in the U.S., often using datasets supplied by the AI provider on their customers or datasets available to the universities, which are also often based on local behaviors and regulatory constraints.
While this might not seem a big issue, we have already seen the impact of having unvaried datasets within the same context and country, with AI systems for justice departments exhibiting bias towards minorities or facial recognition working better for a specific ethnicity as they have been trained more on those types of images.
This means if businesses want their AI to be successful, they must ensure they have the most diverse and broad datasets possible. This includes ones from whatever market you are trying to access, whether Germany, Mexico or India. Only by including this high-quality, healthy data will algorithms be able to understand local nuances and signals that can be crucial in helping a business thrive in that market.
This does not mean companies need to collect their own data in these areas. It is OK for companies to buy rights to existing data as long as it has the necessary information needed to be effective and of course, it meets existing regulatory norms for its safe use. For example, the social contact details for potential targets as, given many workplaces are now hybrid, you can no longer guarantee people will be by the office phone or at their office building.
Up to date information will set you apart
It is the nature of AI that it learns from historical data. From there, it can predict what might happen in the future. This is a crucial part of being able to identify whether a company or business is looking to make a purchase by examining their activity, digital footprints and comparing it to actions done by buyers in the past; information that is crucial for companies’ sales and marketing departments.
However, while this can be effective, it is also important to remember that the world is constantly changing, and predictions based on trends last quarter or a year ago will not be as accurate as one based on the latest known information. As an extreme example, if companies were working off of 2019 data in the second quarter of 2020, then their information would have been wildly off due to the dramatic change to business conditions brought about by COVID-19.
This is why it is vital companies find and have access to relevant information, real world events and datasets that are continuously kept up to date and that data is able to be traced back from its source all the way through to when, where and how the outcome of an AI algorithm is being utilised. This will ensure that the AI is able to make its decisions based on the latest impactful events, not ones that happened a quarter ago, which is usually the length of time before most industries update their databases.
Along with this, by being able to track the data, businesses can ensure it has come from a reputable source, and fingerprint (or audit) the actions of the AI. This is something that is becoming ever more important as different geographies begin to implement new rules and regulations around how data is used and clarity on the explainability of AI outcomes and systems.
By taking these steps, it is possible to create adaptive AI processes that can work or be configured to serve any market anywhere, enabling a business to be as successful in central Europe or Asia as it is in North America.
Shakeel Itoola
Demand Science
