Approach your colleague and ask about his or her thoughts on Artificial Intelligence. About seven out of 10 people would mention that, presumably, the powerful technology may replace humans and dramatically disrupt the labor market. Yes, this is what people have been reading about in media. Despite “AI-powered future” talks hanging in the air, the actual AI revolution remains more of a distant concept.
Despite Artificial Intelligence bringing impressively beneficial insight to numerous industries, from medicine and logistics to all retail domains, there is still one significant blocker to its mass adoption — the prevailing mistrust of the end users. At the same time, AI skeptics are speculating that the era of AI has not really started. Their argument stands to reason, since the mass-scale adoption of AI-driven solutions becomes possible only after the end users fully understand all the processes that led to particular suggestions.
Artificial Intelligence is Coming of Age
The pace of development that AI gained turned out to be both a blessing and a curse. Although AI-based solutions have quickly made their way into a significant number of industries, the human ability to trust and get used to them is staggered. Research from Fayetteville State University in North Carolina confirmed trust remains the major issue with AI, because of the non-transparency of how AI processes information and makes decisions. “To resolve the notorious ‘black-box problem’ it is vital to instill trust in advanced technologies among the end users,” said Sambit Bhattacharya from Fayetteville University.
Speaking of explainability’s importance, it’s worth recalling some of the benchmark AI projects that were once meant to become the disruptors within their domains. IBM Watson and Watson for Oncology started as unparalleled AI-based projects initiated to revolutionize healthcare and related industries. Yet both projects failed due to end-user distrust of the suggestions made by IBM’s supercomputer. They considered the calculations too ambiguous or illogical, and so denied them en masse. Thus, the two projects gracefully demonstrated the truth that remains actual to the day: it is not enough to design a superior technology without it being comprehensive and transparent.
Advertisement
Ever since, explainability has been a prerequisite for AI infiltration across various industries. This is why many businesses have started investing in explainable AI, and retail solution providers are no exception.
Unpacking Black Boxes in Retail
Ultimately, AI has nothing in common with Skynet, Altron, or any other sci-fi concepts. It is just an algorithm that’s better than humans…in calculations. Analyzing gillions of data points, AI algorithms make millions and millions of combinations to help users make the right decision.
The market is flourishing with versatile solutions that fuel supply chains, supermarkets’ CRMs, logistics and pricing. For example, the implications of AI use in various areas of the supply chain range from reducing error rate by 30% to eliminating product wastage, restoring supply chains and optimizing delivery routing. But as with anything else, the issue of AI transparency also remains evergreen here.
To resolve it, software providers de-veil how AI arrives at particular decisions. For example, in logistics, certain clarity can be reached due to a great wealth of data that AI consumes to provide interpretable calculations for the end users. Here, the decisions can be explained because of the known values like weight, transfer costs and distances. Thus, approving AI-based suggestions, in this case, becomes much more straightforward and natural.
In turn, AI solutions for the supply chain struggle with a lack of clean, accurate data to make users trust its suggestions more. To crystalize the algorithms’ work, software providers strengthen the completeness of data via diversifying relevant systems and sources and making the algorithms adaptable to client needs. Thus, end users are notified if the advanced models analyze the systems that embrace the supply network, together with all of the conditions and constraints of its members.
AI’s potential also brings notable results in improving supermarket CRM systems and delivering custom product suggestions based on the analysis of in-store video recordings and customer position data matched to that of the products. Transparency is achieved via a set of clear deliverables like the actual position of a customer, average time spent in front of the shelf and the products picked. The insights, fed into a CRM system and processed with the help of advanced algorithms, result in personalized product recommendations and accurate demand projections for a specific item.
Traditionally, AI decisions that may dramatically impact financial performance arouse extreme levels of caution and distrust. Non-transparency here can sign the death warrant to the very advanced and state-of-the-art solution. In each case, the developers try to solve the black box problem in different ways, but let’s look at an example from the domain where AI has been successfully used for a long time — pricing.
To adequately demonstrate the scope of processes and variables behind final price recommendations, tech providers introduce interpretability features that open up a whole new level of insights to the end users and strengthen their confidence. Thus, users can see how the upper and lower limits are calculated, along with price recommendations within specific price ranges.
AI algorithms consider various econometric factors and business constraints to provide strong reasoning to any given price recommendation. It can be acceptable minimum markup, maximum possible percentage of price change compared to the previous period, seasonality, price index, specific product parameters, competitive data, acceptable percentage of a price change, cost, price thresholds and business metrics to grow or to protect.
In fact, solution providers push the explainability limits even further and offer users a look at the visualization, demonstrating how demand can react to any price changes within fixed limits and goals. Altogether, this information significantly facilitates users’ adoption of AI-powered solutions and opens the way for advanced technologies in retail.
Summing Up
Although AI is taking over various domains of human life, the actual practice has proved an advanced solution is not enough. The total adoption of the technology directly depends on the end users’ trust and the transparency of the processes behind the mysterious AI technology. Solution providers already are working on opening AI’s black boxes to let users understand the machines, and truth be told, they are doing this quite successfully.
Alexandr Galkin is CEO of Competera. He is a serial entrepreneur and tech enthusiast with 12+ years of experience in auditing and consulting global retail enterprises. Having successfully started and sold an outsourcing company, Galkin founded Competera, an AI-powered pricing platform for online, offline and omnichannel retailers, which grew into an acknowledged technology company with about 100 employees on board. When not engaged in fixing the broken pricing world, Galkin is a frequent contributor to high-profile retail and business media.
