What are the biggest barriers to AI adoption for retailers?

Discussion
Photo: Getty Images/Jiraroj Praditcharoenkul
Feb 24, 2020
Tom Ryan

According to KPMG’s “Living in an AI World 2020 Report,” retailers have some optimism, some skepticism and some pessimism about how artificial intelligence (AI) will impact the industry.

The study explored how 751 insiders across five industries, including retail, view the future of AI in their sectors.

On the downside, 64 percent of retail insiders agreed that the use of AI to help businesses is more hype than reality right now.

The study also identified numerous challenges retailers believe they face in capitalizing on AI’s potential:

  • AI readiness: Just 43 percent of retail respondents believe their employees are prepared for AI adoption. Relatedly, only 52 percent say their companies offer any type of AI training.
  • Job loss threats: Respondents believe only 26 percent of retail employees are supportive of the adoption of AI, partly due to concerns over job loss. Sixty-two percent believe retail workers are worried about AI taking away jobs and 54 percent are worried that their own jobs could be replaced by AI someday.
  • Data security: Seventy percent of retailers said perceived threats involving consumer data security and privacy may slow AI adoption. Ninety percent agreed that their companies need to be responsible for implementing a code of ethics.

Despite concerns, 80 percent of retail insiders say AI technology — such as chatbots and self-checkout — is already regularly being used to alleviate customer service issues. Eighty-six percent believe AI has the potential to significantly improve organizational efficiencies.

Among specific applications, customer intelligence will see the biggest impact within two years from AI, say 56 percent. That’s followed by self-checkout services, 55 percent; chatbots for customer service, 45 percent; supply chain planning, 44 percent; and marketing/advertising, 43 percent.

Bill Nowacki, managing director, decision science, KPMG, believes AI will prove to be particularly beneficial in helping retailers fine-tune execution at the local level. “There’s a push to say, Can I get better with local relevance and placement? Am I in the right locations? Is my format right? Do I have the right items in the store? AI is really helpful in all these areas,” he said.

DISCUSSION QUESTIONS: Where do you see the pain points, perceived risks and challenges facing U.S. retailers related to AI? Which AI benefits are realizable now and which may be more hype than reality for many years to come?

Please practice The RetailWire Golden Rule when submitting your comments.
Braintrust
"AI needs to be implemented with great care by retailers so as not to undermine the only true advantage a brick-and-mortar store has over e-commerce: human touch."
"...inserting new tech in a complicated supply chain is a lot harder than adding a chatbot to a website."
"As with any new technology, AI adoption rate is impacted by faith. Faith that the AI can provided an improved result or, more importantly, can make life easier for the user."

Join the Discussion!

23 Comments on "What are the biggest barriers to AI adoption for retailers?"


Sort by:   newest | oldest | most voted
Shep Hyken
BrainTrust

AI is a broad subject with many different uses/applications in the retail industry. Self-service checkout and chatbots are very basic customer-focused uses of AI. Targeted advertising and personalization comes from data mining. Managing logistics and supply chains are yet more uses. The key is to know exactly what you want to achieve. If it’s customer-facing, be careful not to become so enamored with technology (AI) that it takes the human element out of the experience to the point of alienating customers. If it’s behind the scenes, the application must make things easier — not more complicated. If there were one word to some this up, I’d say it would be “balance.”

Dr. Stephen Needel
BrainTrust

AI is long on hype and low on delivery (as separate from automation, which if done well can be long on delivery). Fine tuning execution is not an AI problem, it’s an organizational problem or a research problem.

Suresh Chaganti
BrainTrust

Marketing and merchandising are easier places to start from an adoption perspective. Send highly relevant, personalized promotions, and make sure the assortment and planogram decisions are completely data-driven. This will get the most out of existing investments.

The resistance is highest where employees and customers are involved. That means robots in the warehouse and self-service kiosks in the checkouts will be adopted relatively slowly. The reasons are obvious – concerns about job loss, frustrating experiences, privacy issues. Change here demands behavioral shifts and changes in customer psychology, which will be slow in coming.

Ralph Jacobson
BrainTrust

Much of the hesitation and perception of “AI hype” is driven by those putting the cart before the horse. Don’t look for where to stuff AI into your systems. Instead, continue to address business challenges. And where those challenges may benefit from a specific capability (e.g., natural language processing, etc.), that is where to investigate the myriad AI APIs to see if they will generate tangible results and ROI.

Michael Decker
BrainTrust

AI needs to be implemented with great care by retailers so as not to undermine the only true advantage a brick-and-mortar store has over e-commerce: human touch. Routine services such as automated checkout should be used to ENHANCE personal service by freeing up employees to improve the one-to-one connection that great sales help can deliver to a customer that needs help, opinions, and advice in their shopping adventure. Once the novelty wears off, nobody wants to talk to robots.

Ken Morris
BrainTrust

I think a big problem is the definition of AI. Some definitions seem to essentially be “AI is whatever hasn’t been done yet.” So what exactly are we talking about? Self-checkout is a process and to me not AI, to give one example.

The biggest stumbling block to real AI in retail is the legacy applications that support retail today and their siloed architectures which make integration with natural language, robotics and machine learning a daunting challenge. Retailers need to create unified platforms that make this transition possible rather than trying to satisfy the need for hype with robots that sit on top of tech silos and just roam the aisles with little functionality.

Georganne Bender
BrainTrust

I agree that currently AI is more hype than reality, but I also believe it has a place in retail, especially to supplement when store associates are not available or the task is novel or best done by AI.

I’m curious about the comments in the article that AI can help retailers fine-tune execution at a local level. Local relevance and finding the right format, location, and items for a store seem more like spending time in a location decisions. Rachel Schechtman just demonstrated that with Macy’s The Market in Texas.

Lisa Goller
BrainTrust

Data integrity is the top pain point and opportunity for AI. Retailers need to smash internal silos that prevent data from flowing across functions. They also need to ensure their data is accurate and up-to-date for successful supply chain collaboration.

Also, labor represents a massive opportunity to find efficiencies since retail is such a people-centric industry. AI is already helping retailers streamline recruitment and scheduling processes to save time and minimize the risk of manual data errors.

Over the longer term, AI will create e-commerce benefits for a competitive edge, including:

  • Efficient product discovery through online visual search and effective recommendations;
  • Relevant personalized marketing thanks to rich consumer insights;
  • Effective online pricing and assortment adjustments to reflect price elasticity and demand.

In addition, as a recent RetailWire discussion found, the use of AI for voice shopping will still take time as consumer trust needs to outweigh the technology concerns.

Meaghan Brophy
BrainTrust

I find it hard to discuss AI in one broad bucket because it has so many different applications and use cases. Like the study suggests, chatbots, self-checkout, and even personalization are practical and low-barrier uses for retailers. They are relatively easy to implement, and the benefits in improved customer service are tangible. AI can also be beneficial for predicting supply chain demands (Amazon has been doing this for a while). However, inserting new tech in a complicated supply chain is a lot harder than adding a chatbot to a website.

Peter Charness
BrainTrust

AI has to be seen through a different lens than the typical IT project. Implementation times are more relative to understanding data and training models, not installing software. This is a process that for some retailers will take a considerable period of time — months if not years. Furthermore, once implemented models need continuous monitoring and retraining so, unlike a typical solution, the implementation is the start or maybe the middle of the work effort not the completion.

More challenging for many retailers will be the recruiting and retention of data scientists able to understand the business and provide this long term care and feeding. AI has great potential, but it’s not a magic pill that can be swallowed for immediate benefit.

Oliver Guy
BrainTrust

As with any new technology, AI adoption rate is impacted by faith. Faith that the AI can provided an improved result or, more importantly, can make life easier for the user.

Management of expectations is key – there may well be an expectation of perfection immediately and while an algorithm may get 99 challenges right, it is human nature to focus on the one it got wrong. Therefore it is important to remind teams of this when promoting adoption. Being able to operationalize learnings quickly and re-deploy is likely to help with ongoing improvement and faith building.

Harley Feldman
BrainTrust

The biggest two points are accuracy in store inventory and customer intelligence to understand real demand. Very few people understand what true AI technology is so they are implementing rules-based engines which are better than nothing but do not provide the benefits of technologies like machine learning which require a higher skill level. The big benefits to retailers will only occur when advanced technologies can be utilized. Getting in the way is the accuracy and granularity of data collected and the skills required to implement such AI.

Steve Dennis
BrainTrust

For the most part AI is still more hype than reality for most retailers, but I do see it gaining traction.

It’s important to remember that AI is generally pointed at two impact areas: efficiency (i.e. speed of service or taking out costs via automation and the like) or effectiveness, by enhancing the quality of service or improving customer relevancy (often through improved personalization). The former is getting the most traction.

The effectiveness side is largely limited by many retailers not having the basics of a really good customer insight and personalization strategy in place. Throwing money at more sophisticated solutions often doesn’t work if you don’t have the foundation in place. First things first.

Mohamed Amer
BrainTrust
The biggest barriers are going beyond the hype and getting the c-suite to understand what artificial intelligence is, and how it can be applied to drive top line growth and generate bottom line efficiencies in their specific business model. Even more impactful and profitable is the potential to change your existing business model through novel application of AI. However the necessary mindset here is not one of incrementalism in the sense of continuous improvement, but one of pursuing new ways of doing a process (simplifying) or even completely eliminating processes (eliminating friction). Artificial Intelligence has such a potential, but it doesn’t just happen by flicking a switch. As AI gets bandied about its luster is well deserved, yet the hyperbole makes it easy for reluctant executives and analysts alike to snicker and disregard. Any change that challenges the existing assumptions upon which the business is founded, or by which past successes were achieved will be considered safe to ridicule and ignore. In the 2020s, Artificial Intelligence will be as ubiquitous and as useful as the… Read more »
Gene Detroyer
BrainTrust

I have a problem with the sentiment that AI is “more hype than reality.” AI is already quite successful in manufacturing, logistics and supply chain, Intelligence and security, and healthcare. It will only get better — and at lightning speed.

The impact at retail is not about self-checkout or funny robots running around the store. It is about everything in the business from the manufacturer to the customer taking home products. It is about efficiency and ideal targeting. Per today’s discussion about shelving, AI will provide the optimum apportionment and location for each individual store.

The biggest pain point is that retailers are thinking small (funny robots) and not understanding how similarly effective AI can be at retail as it is proving in manufacturing, healthcare or autonomous vehicles.

Doug Garnett
BrainTrust

AI adoption won’t be visible — most employees won’t even notice. And that makes this topic even more critical.

I highly recommend the new book by Melanie Mitchell of Portland State and the Santa Fe Institute — “Artificial Intelligence: A Thinking Guide for Humans.”

Key in this book is discovering how relatively simple it is to fool an AI-based system. Also how easy it is for an AI deep learning system to embed prejudices — not clear them up.

AI is merely a new set of algorithms — only this time humans don’t drive every step but teach the algorithm things in such a way that we can never know what it learned.

As a result, Mitchell gives excellent examples of the mis-training of AI and how hard it is to detect that it has happened.

Retailers should embrace AI — and do so cautiously, knowing that they don’t control it. And if they implement it poorly, it will end up controlling them.

Gal Rimon
Guest
Much of the discussion around AI centers on this technology in isolation. That AI operates on its own. It takes over tasks and replaces people. I’m not about to enter that debate. Rather, and especially as we’re talking about applying it to the brick-and-mortar environment, we need to look at AI operating in a collaborative manner with HI, Human Intelligence. As amazing as it may be in crunching numbers to arrive at data-based predictions, AI cannot look beyond those numbers. Further, it doesn’t have innate creativity (even after lots of machine learning). And it doesn’t have empathy. Those are distinctly human qualities. When you draw on an HI+AI construct, you can arrive at decisions faster but, with the human element, they’re better. Beyond that, and of considerable importance, HI+AI will allow store/department team leaders to have more time to develop their sales teams because a host of administrative tasks (which can eat up more than half a work day) can be taken over by AI. To be sure, AI is not a villain per se.… Read more »
Gib Bassett
BrainTrust

I find it hard to believe that “43 percent of retail respondents believe their employees are prepared for AI adoption.” I’m not sure many people understand what AI is, represents or the adoption challenges. There is a lot of focus on the use cases (chatbots, robots, self service checkout), as opposed to developing AI as a competency, a muscle to be developed and maintained. As more companies adopt point solutions around specific use cases, any efficiency or effectiveness gains becomes table stakes. Which, in effect, nullifies the ultimate value to be attained through an executive-led initiative to become data driven and prioritize analytics as a core competency. Of which AI is a component.

James Tenser
BrainTrust

AI is beginning to shed its buzzword status, as more retailers put data scientists on the payroll. On its own AI is a hollow concept. In practice however, it can be the core of a host of new or improved operational and experiential retail practices.

Hype tends to take over where more nuanced understanding is in short supply. So it is with artificial or machine intelligence. Learning systems present certain potential advantages compared with systems that operate based on human-designed algorithms.

Retailers need to make the conversation around AI be more specific — natural language systems, for example have very different purposes compared with image-recognition systems, automated analytics, or supply chain optimization systems.

Starting to wince when I see images of robots used in this discussion. Here’s a tip for the wary buyer: the more humanoid they look, the less intelligent they actually are.

Jeff Weidauer
BrainTrust

The two primary questions retailers need to answer before implementing any AI system:

1. Is this solving a problem?
2. If yes, is it a retailer problem or a customer problem?

Quite often, neither of these questions are asked nor answered in the rush to get “something” out there.

Ananda Chakravarty
BrainTrust
The retail industry has been talking about AI for the past few years and we have genuine applications for it, from demand forecasting to online predictive recommendations. However, two big barriers stand in the way — data infrastructure and public understanding of AI. AI is data heavy, so without the infrastructure to support strong, clean, and accurate data, AI cannot perform. In some cases it’s almost impossible to secure relevant data that will help an AI system solve an app problem. Public understanding on what AI is doing is the second big barrier. AI spending continues to rise, so adoption is increasing, but consumer facing apps don’t highlight the AI pieces nor is it easy to understand underlying algorithms, neural networks and specifically what kind of performance lift the AI components provide. So long as we are challenged to measure, there will be skepticism and hype. Sometimes the data scientist needs to show the benefits that their new AI system has delivered on to impress the exec team, which still sees a black box. As… Read more »
Kai Clarke
BrainTrust

We first have to define what AI is and how it will be applied at retail. Just throwing out the word AI is too broad and indiscriminate. The AI areas touched upon in the article are basic information or cashier assistance points, and more mechanized or communication robots. These are still just being identified and growing at retail, since they are in their infancy. We have to wait and see how AI matures and is applied by retailers as they better compete for the consumers attention.

Andrew Blatherwick
BrainTrust

I’m surprised that they are surprised that 64 percent of retail insiders say AI is currently more about hype than reality. This coming from KPMG, who are part of the cause of this, along with their other consulting friends, by hyping AI as the solution to everything.

They don’t realize retailers live in the real world not the test tube world of the consultant, who can spend time reading and understanding this stuff but have NO risk whatsoever if it goes wrong because they can walk away from an implementation if it fails. Retailers are realists who live in the real world and have busy jobs to do – they will not take risks at the bleeding edge until it is proven to work.

wpDiscuz
Braintrust
"AI needs to be implemented with great care by retailers so as not to undermine the only true advantage a brick-and-mortar store has over e-commerce: human touch."
"...inserting new tech in a complicated supply chain is a lot harder than adding a chatbot to a website."
"As with any new technology, AI adoption rate is impacted by faith. Faith that the AI can provided an improved result or, more importantly, can make life easier for the user."

Take Our Instant Poll

Which of the three areas identified in KPMG’s survey do you see as the biggest hurdle slowing adoption of AI?

View Results

Loading ... Loading ...