Artificial Intelligence and machine learning provide insights at a level of granularity and speed that we would not have dreamed of even five years ago. For some CIOs and business leaders there is a perception that taking humans out of the decision-making process will offer resource savings, speed advantages, and substantial commercial benefits. But in the dynamic world of retail, the key to unlocking the insights that AI provides remains people—smart, experienced, data-savvy people.
Retailers capture a vast amount of data and this can be used to provide a clear view of how customers behave through the shopping experience: trip cadence, favorite or substitutable products, price and promotion thresholds, and even, for more sophisticated retailers, predicted impact of marketing activity and competitor actions. Initial use cases for AI have frequently been to improve demand forecasts, modeling better assortments, promotions and prices as well as more effective communications. AI is also being used for personalization learning what offers work for individual customers in any given situation.
We are still seeing that the most impactful factors to include in AI models, remain those that come from within the retailers own data. However, ever cheaper data storage and improving processing power encourages the creation of ever bigger data lakes and the inclusion of even more disparate data sets. Of course, AI programs can tie these together to identify new and (sometimes) relevant characteristics but caution needs to be applied as there are diminishing returns in driving sales.
Where data relevance and impact can be determined, collaboration between organizations to add new data points is an emerging trend. Using this additional information smartly, retailers can create: store experiences, develop promotions, and communicate to customers on an even more personalized basis. As long as privacy concerns are managed and organizations are transparent and appropriate, customers are willing to give permission to aggregate their data/ information as long as they feel they are getting value in return.
The challenge remains that information and insights from AI needs to be operationalized, it must be funneled to drive the right, business actions and outcomes. Insights, especially from AI, are raw objects that exist without full situational context, such as a list of products that could be discontinued. The data and analytics provide information and insights for consideration, but it’s ultimately up to the real intelligence of a human brain to decide how to proceed. We have numerous examples of damage done to category sales when items that had strong niche appeal, marketplace uniqueness, store-specific strength, or cross category appeal were delisted. Category managers can’t afford to make mistakes and are time poor. They need standardized reports, familiar metrics and well-designed business processes, to make sense of the AI outputs without losing speed or having to be retrained in advanced analytics. But, they do need to know a little about why the data is saying what it does; to be effective AI must not be a completely black box solution.
Retailers already benefit from AI programs and as the programs become even more advanced, such as filtering though unstructured data like social media posts, the conversation will become even more personal. Product suggestions will move from associated purchases to diet-specific or taste-specific recommendations.
We believe the next phase will be when the AI conversation turns on its head and the customer starts to lead. Eventually a customer will be able to create a shopping list, and their AI will make recommendations based on their dietary, economic, environmental, brand, etc. preferences. Search available retailers find the best prices, and have the products delivered to their door. The retailers who thrive in this world will be those who do not just react but identify opportunities in near real time and that will depend on people, not just AI. Are you ready for this?