Getty Images/josefkubes
Why do so many people working in retail find AI scary?
While artificial intelligence (AI) isn’t new — it’s been around since the 1950s — it is a hot topic right now. This is primarily because we now have the data and the computing power available to use it efficiently. Today’s retailers can take advantage of AI to streamline and improve business processes, but there’s also a lot of misinformation around the technology that must be addressed.
I’m not surprised, for example, that retail staffers are concerned that AI could affect them negatively and endanger their jobs. According to a recent KPMG study, only 26 percent of retail staff are supportive of the adoption of AI. And that is just one example — the press is full of articles implying AI is the great problem solver that will take over the need for human thought and intervention across the board.
This is simply not true. Because it’s a common misperception and fear, however, we need to shift the conversation away from AI and customer service (think robots in the aisle, ready to talk to customers and restock shelves). Instead, we should focus on how AI can be used to gather and interpret data and how those results can enhance — not replace — existing retail processes.
Retail is largely about managing the flow of inventory through the supply chain. Incorporating AI into demand forecasting and inventory optimization can improve product availability, reduce waste, smooth the flow of goods and improve operational efficiency in distribution and in stores. It can help produce better demand forecasts that accurately reflect the impact of promotions, cannibalization and even changes in the weather. It can also optimize orders to minimize spoilage while avoiding out-of-stocks and ensuring controlled ramp-ups and ramp-downs when a retailer’s assortment changes.
Plus, when AI is folded into workforce optimization software, schedule creation becomes less time consuming and much more accurate. Retailers can quickly develop work schedules that support both business initiatives and human needs, using forecasting data and considering local labor laws and individual work requirements. With AI to assist this process, the stress of either over or understaffing can be nearly eliminated, while offering store associates more predictable working hours.
AI really shines when it is incorporated into software to gather data that can be used to improve existing jobs and processes and, in turn, the lives of the people involved. Properly delivered, nearly anyone — not just mathematicians and data scientists — can interpret and use this data to move forward a project or improve a process.
What is the future of AI over the next 10, 20 or 50 years? I don’t think anybody knows at the moment. You can speculate, but I just don’t see it as a substitute for people. I believe those in retail should ask themselves, “If I want to further my career, how do I embrace AI and get the best out of it?” It shouldn’t be seen as a risk; it should be seen as something that helps you, not threatens you.
The good news is that, according to KPMG, “eighty-five percent of retail insiders believe AI has the potential to significantly improve operational efficiencies.” Adoption in the industry, however, is slow because of all of the barriers I’ve noted. What can we do to improve the perception of AI and get more people on board?
Discussion Questions
DISCUSSION QUESTIONS: How is AI functionally changing retail operations today? How will it functionally change retail operations in the next five to 10 years?
Poll
BrainTrust
Bethany Allee
Senior Vice President Marketing, PDI
Doug Garnett
President, Protonik
Change is scary and when it’s coupled with highly disruptive technology, the unknown is amplified. AI functionality is currently helping retailers better understand and cater to customer buying needs. AI helps retailers automate simple tasks. This allows them to deploy their human workforce in a way that delivers better service to customers. In the next 10 years, AI in retail will focus on better understanding customer buying habits, and customer service will become an innovation platform.
When we talk about AI and retail, the common image is one of C3PO stocking shelves and helping shoppers. But the real value of AI will be to assist retail workers in doing their jobs more effectively, not to replace them outright. It’s up to the industry to change this misperception if it wants to gain support from internal teams.
ATMs reduced the number of tellers, self-checkout reduced the number of cashiers. It’s no wonder retail employees are looking at AI askance. Automation has a habit of disrupting more than commerce; it has disrupted jobs.
Better is education about which jobs AI will help make easier. Still, I’m sure many will believe it when they see it.
Today in retail, AI is being used to help with labor scheduling, task creation and assignment, shelf edge inspection for out of stocks, pricing compliance, planogram compliance, hazard and spill detection, freshness and waste mitigation, and plenty more. However, it is still early days. Inherently, “AI” is dumb. Its view of the world is the distribution of the data it was trained on; fall out of distribution and watch how AI can go bad. I think that within the next 10 years there will be more causal models, then correlation-based, and real-time human preference seeking, which will lead to even more powerful machine agents.
AI is still very much a nascent technology. There is always a reluctance to adopt change and AI introduces new variables where retailers aren’t sure of their place, about employee replacement, about reassessing knowledge of the trade that has stood for centuries. The changes we will see now are nothing like what can be expected five to 10 years in the future. We are seeing improvements in the supply chain, in demand forecasting, and to a lesser extent in predictive modeling for customer personalization. Applications with real ROI will take precedence, but retail is still discovering these. The future 10 years out will have far more complexity but the reliance on AI will be more normal and more invisible, where it will be part of the toolset most retailers use.
Everyone talks about the great benefits of AI at retail. I agree that these benefits are there, but the fundamental challenges of implementing AI with a retailer remain: the cost and people. Any retailer in the U.S. has more than enough data stored to implement an AI program to support and improve their business. What many retailers don’t have is the capital to spend on the implementation of AI in their business. Beyond AI, capital cost is the cost of employing the right people to make AI actionable in the retail space. Until the retail business can overcome these two issues, AI will remain a non-starter for many retailers.
Tesler’s Theorem says “AI is whatever hasn’t been done yet.” When I think of AI in retail, I really think of BI (Business Intelligence). I like how Gartner defines the BI space into four phases:
That prescriptive capability is what I believe AI offers in retail. What action do I take about what happened? Learning what data constitutes a high likelihood that a theft has occurred at a register is an example of this. Tracking something like post voids at POS and highlighting anomalies that indicate likely theft or training issues and launching a closed loop case via email or text with notification and escalation, is a great use case. These type of opportunities exist across every retailer whether it is a toilet paper shortage or cleaning products out-of-stock as it is today. There is a mission critical challenge to improve process by leveraging AI. This is a tool that should be embraced rather than feared and we will see much more AI/BI in our immediate and long term future.
AI is already pervasive in retail – and the pandemic will only accelerate its adoption.
Currently, AI helps with:
The pandemic is increasing the volume of e-commerce data as more consumers move online due to self isolation.
To avoid drowning in all that data, more retail companies will invest in AI to spot patterns and make data-driven decisions to boost productivity. AI can free up time spent on repetitive tasks like data entry and number crunching so companies can reallocate effort toward serving consumers and streamlining operations.
Over the longer term, we will see more use of AI for virtual assistants for voice commerce, and checkout-free service like Amazon Go stores.
AI, as an overarching technology, will definitely replace some repetitive jobs, but will also provide many more jobs to maintain and develop those technologies. Remember, not all that long ago we didn’t have data scientists, for example.
I will go off topic a bit to try to give an example of the magnitude of AI. The University of North Carolina has used AI to absorb hundreds of thousands of pages of medical information. Perhaps thousands daily — an aggregate amount of pages in a short time beyond what any single professional could read in a lifetime. Someone then feeds a patient’s symptoms into this. The system has about 80 percent more accuracy (and operates much quicker) than the human professionals for diagnosing ailments.
Now imagine that capacity being applied to every aspect of retail from customer interaction to a global supply chain. Does it make the whole system more efficient? I can’t believe it won’t.
Retailers historically have been reluctant to embrace new technology and trends. Those who do will generate a very, very wide gap between themselves and the laggards.
Will some people’s jobs be eliminated? Absolutely! But as with all fearful technology developments in history, the new technologies have, in total, produced even more jobs.
Count me among the 85 percent! AI is already making its presence known in the retail supply chain and is quickly expanding into operational processes and to some degree customer-facing initiatives. Workforce automation isn’t something retail employees should fear, on the contrary, it will lead to significant efficiencies that will make their job easier. Repetitive tasks will either be eliminated or be handled much faster than before, leaving store associates to spend more time in higher-value activities with customers. In terms of in-store operations there are many areas that will benefit, including labor scheduling and many other repetitive tasks. Customers will benefit from capabilities like AI-enabled visual search, for example. I expect we will see many areas ramping up their use of AI as a result of the COVID-19 crisis. Not in the form of robots that replace workers, but in the form of AI-enabled tools that simplify many tedious functions and make the job a better, more enjoyable, more productive one.
AI doesn’t replace what people are doing today, it enhances their ability to do what they should be doing but don’t have the time or energy to execute. Localizing assortments, for example, is something most retailers know they should be doing, and want to do. It’s an order of magnitude more work than the teams are currently capable of — that’s why AI should not be viewed as scary or a replacement for people.
At this point in time, I think retail workers fear for their jobs due to COVID-19 more than any AI.
To answer this question, if retail workers fear AI it’s because their leadership has not communicated a strategy for how AI and analytics supports the business. People drive processes, processes are improved through AI, and you will not get adoption and value if your people are not clear on the strategy. All the research into what works/does not validates that this is what happens in cases when companies report failed AI initiatives.
Collective intelligence is a good way to frame how people and machines working together create better outcomes than they can individually. In cases when jobs change due to AI, it is often because repeatable tasks within a specific job can be automated. The job itself is not lost, but changes becomes more efficient. As a result, you may need fewer new people to do the “old” job, but more people to do the “new” job that leverages human talents that AI cannot replace.
We SHOULD be cautious about AI. Reading Melanie Mitchell’s excellent book about AI, there are incredible gaps in it and incredible lapses in what it can/can’t do. Unfortunately, it’s very hard to predict where those gaps are. Even more concerning, it’s possible to game an AI system.
AI systems have some tremendous potential and should be used for that potential. But they aren’t intelligent — they just offer new ways to get to answers. And unfortunately, these new ways quite often don’t have any way to get inside the algorithm to understand that choices that are being made — and hence the errors or risks.
As Mitchell observes, AI systems which identify what’s in a photo, for example, never understand the photo. So when they make mistakes, they make very big mistakes.
AI is a buzzword for most people, who for the most part have no interest in understanding it, usually exhibited by a glazed look. Face it, Artificial Intelligence is a conversational non-starter. A bag of misrepresentations combined with employee fears of unknown outcomes.
Education is key. Flip the fear of AI into Employee Subject Matter Experts training machines to improve operational efficiencies. With AI, employees do get to think. AI is the dark horse. Retailers who educate their employees on the opportunities of AI will be the new power players of retail. Watch out Amazon.