What companies need to know before using AI

Photo: @kirsty via Twenty20
May 06, 2019
Gib Bassett

While most companies struggle to make sense of artificial intelligence (AI), some have made significant progress. Within retailers and consumer brands, two reasons for this separation are bubbling up to the surface: “explainability” and having the capacity to test many use cases.


Retailers and brands should not embark upon the use of AI without first understanding how the insight is generated, what data sources it relies upon and the ethical implications. Why? Those in the business will definitely ask, your customers may ask, and ultimately the entire company could be called to account to explain why a particular action was taken. It could be as seemingly benign as a dynamic price offer for a customer — an action a large retailer recently found was not so benign.

Consumers are becoming increasingly anxious today about their data privacy and how their data is used. This is only going to escalate and those prepared to answer for it will be better off. So business, technology and analytics staff need to be educated consumers.

Test and Learn

The ability to test many use cases goes hand-in-hand with explainability. Why? Analytics, like machine learning (behind most AI use cases), are all about continuous improvement. If you can’t understand the most basic factors driving an AI solution, how can you hope to measure and improve? To scale value and reap the most benefits, companies of all sizes need the capacity to evaluate, improve and completely renovate AI use cases.

When you can do this, you support efficient exploration of many use cases that map to your company’s business strategy. Companies can derive greater benefits when AI flows from a use case portfolio as opposed to leaving it to silos to explore point solutions in a disconnected manner.

The companies that address these two points well are among the largest and most experienced with advanced analytics. The challenge for the masses is how to adopt these best practices with less human, financial and technical capital. Anything short of a thoughtful plan supported by the CEO is probably going to fail.

DISCUSSION QUESTIONS: Why do retailers seem to have a hard time explaining their use of AI and fail to adequately employ testing and learning strategies around the technology? What should retailers do differently to address these challenges?

Please practice The RetailWire Golden Rule when submitting your comments.
"A lot more care and consideration needs to be put into introducing and rolling out all new technologies in the retail space."
"the more effective approach is to take a look at your prioritized business challenges and see which processes could improve via augmented capabilities."
"I’m one of the first people to embrace a new technology, but I gotta say – people need to be way more careful about AI than they are being."

Join the Discussion!

18 Comments on "What companies need to know before using AI"

Sort by:   newest | oldest | most voted
Cathy Hotka

There are a lot of retailers who are already using AI, whether they know it or not; it’s embedded in a number of modern software products. The more fundamental issue is whether retailers can leverage the oceans of data they have, in order to transform relationships with customers.

Dr. Stephen Needel

Retailers have a tough time explaining this because mostly they don’t need to use it – they are buying into it because it’s the hot term. That doesn’t mean AI can’t find some interesting things to try, but so could smart analysts and you need them anyways, as Mr. Bassett says, to understand what the model really is. Keep in mind most AI models are indecipherable as a mathematical equation, meaning you can’t play with purchase drivers to see what happens – you have to take it on faith that the model is right. Never take on faith that a model is right.

Neil Saunders

As with most technologies, a lot of retailers simply jump on the AI bandwagon without fully understanding it. Moreover, they don’t think through the purpose of AI or what impact it will have on the business. A lot more care and consideration needs to be put into introducing and rolling out all new technologies in the retail space.

Nikki Baird

I’m one of the first people to embrace a new technology, but I gotta say – people need to be way more careful about AI than they are being. The impact is more dangerous in other industries – medical diagnostic AIs that start using incidental, rather than causal factors in making a diagnosis, or AIs that identify likely crimes using nothing better than the biased profiling shortcuts that humans make today. In retail, AI that you can’t explain, and that is not supported by a robust dataset that explores all possible scenarios, to make sure the model really learns what is possible, can lead to discrimination rather than the personalization it’s trying to achieve.

AI is not a silver bullet. It’s not as sexy as the advertising makes it out to be, and if you don’t do it well, you are codifying biases that you can’t identify or control against. And that’s ultimately bad for everyone. Retailers need to temper their enthusiasm, and take a cautious and controlled approach.

Charles Dimov
First of all, we don’t have AI. We have ML – which is machine learning. AI suggests some form of higher cognitive ability. A danger of our current state is that we are in the Wild West. Too many vendors are tacking AI to their brochures, suggesting that this is the ultimate technology. Sadder yet is that many retailers are falling for it, thinking that if they don’t have AI/ML then it just isn’t the latest tech and that they need AI – even if they don’t understand how. ML should be part of your considerations when you figure out where it is most important. Beyond that, you want a system that can explain why the decisions it has taken are the right ones. It is not enough that the computer stated that routing from a warehouse in Siberia is the best option for an order delivery to Michigan – rather than routing from the DC in Denver. It might be the case — but you need to understand why and be able to defend… Read more »
Carol Spieckerman

I haven’t found that retailers universally struggle with explaining AI (slow adoption of it is another topic). In fact, I spoke with a couple of retail execs recently, both of whom provided nice overviews of their AI plans and progress. Retailers that aren’t as confident probably wouldn’t be as vocal and I agree that the main explanation requirement applies to internal stakeholders. They need to understand how AI impacts their roles and how data guardrails will work. This presents a major opportunity for retailers’ data (and AI) partners to simplify features and benefits. After all, retailers are not building and operating AI solutions entirely in-house.

Brandon Rael

The modern buying office or back office are already leveraging AI solutions, which are a part of quite a few integrated solutions. Today’s retail and manufacturing corporate offices are culturally ready and willing to make more intelligent, insight driven decisions, with the support of the power of AI.

With that said, AI is an enabler and tool to help make more predictive analytical decisions. However, retail magic is all about the blending of the art and sciences. Intuition and experience combined with leveraging the power of AI is the winning combination to improve your operations, drive additional revenues, and experience sustainable growth.

David Weinand

AI is hard to explain as it has turned into a marketing catch-all for many of the software vendors in the industry. Retailers have to cut through the marketing buzz and look at specific use cases where AI can provide real value and then seek (and test) solutions that prove that the technology does in fact learn and provide continuous improvement. AI has a lot of potential and will mature to show real benefit. Retailers that are on the front edge of innovation just have to test and learn more.

Ryan Mathews

With notable exceptions, retailers are generally not in the vanguard of any technological revolution, and that is certainly — at least on a conscious level — the case with AI. Cathy is right that most retailers are using more AI than they think they are, but that is passive use at best. In terms of active engagement with AI, I think it goes back to a general reluctance — again with notable exceptions like Walmart — to test technologies. Why is this the case? Because one rarely reaches the CEO or COO position in retail organizations from the CTO and/or CIO slot. So all technologies seem a bit foreign and the ones like AI that get such broad and often contradictory press seem the most foreign of all. If you want to built a technologically savvy retail organization you might want to think about having a technologist run it.

Camille P. Schuster, PhD.

Many companies are adopting something they are sold as AI without fully understanding what it does, how it works, and why they are using it. If top management doesn’t understand AI and if the people creating marketing strategies do not understand how AI is being used, companies can not explain the use of AI well to consumers. Using AI is more than just adopting the use of new software and needs to be viewed as such.

Joan Treistman

I think Nikki Baird expresses the real challenge of using AI. The user must first identify causal relationships and labels that are to be addressed. There are many failed attempts (perhaps outside that particular retailer’s context) that can be reviewed to help guide planning and development. But as many have suggested, AI is really a sexy buzzword and machine learning is often the actual product and not AI at all. Retailers should deal with the potential use of AI as with any major investment. First identify what you want it to do for you, determine if it can do it and who is accountable if it doesn’t fulfill the objectives.

Ralph Jacobson

Many retailers look for types of AI to implement, while the more effective approach is to take a look at your prioritized business challenges and see which processes could improve via augmented capabilities. As an example, if your call center is taxed and becoming less effective at satisfying shoppers, investigate automating the human piece with augmented intelligence in a machine learning chatbot. Or if you are uncertain of your brand perception, check out automated web crawlers, sentiment analysis and customer experience analytics — all using true AI.

Cynthia Holcomb

Bottom line, if a retailer can not explain AI then the retailer does not understand AI. AI is worthwhile only if solving for a relevant solution. Retailers, especially at the C-level are learning quickly. Computer and data scientists are iterating over and over again to find a problem, rather than a solution. The retailer needs to be involved. It’s difficult to create delicious recipes without understanding the art and science of food. True AI retail business solutions will emerge only when a few visionary retailers apply their time and retail knowledge to investigating meaningful use cases of AI to solve for specific retail business solutions.

Shep Hyken

There are many ways to use AI, some of which is how the retailer and customer communicate. And, then there is data collection. It may or may not be obvious to the customer how AI is being used. If the retailer approaches it the right way, the customer should see it as an opportunity for a better customer experience. And, as long as the company doesn’t abuse the relationship, the connection between the retailer and customer can be stronger.

Ken Morris
Ken Morris
Retail industry thought leader
1 year 24 days ago
Explaining the algorithms and math underlying AI is not realistic, but explaining the source of data inputs and the objectives of the output is important to get acceptance from retail executives and the practitioners that will be relying on AI to guide decisions. With AI, retailers can reassess models and reevaluate the data, all without the intervention of a human. AI is able to test and retest data to predict every possible customer-product match, at a speed and capability no human, or team of humans, could possibly achieve. The result is far more accurate decisions. However, don’t forget the concept of “garbage in, garbage out.” That old phrase is as appropriate today as it ever was. Systemic bias and bad data needs to be scrubbed out so neural networks will apply the proper framework of algorithms to make AI unbiased and accurate. Retailers need to be aware of the potential mistakes that AI can make based on historical data and biases and monitor the results of AI to help catch faulty logic. Humans are not… Read more »
Ananda Chakravarty

I’ll share my offsite blog with the RetailWire audience to answer this question – feel free to partake or not. I’ve condensed some key points I observed about AI below. I believe the 8 realities of retail AI are:

  • It’s early days
  • Digital to ops
  • Many applications in retail
  • Invisible to the public
  • Not always ROI based
  • Data dependent
  • Talent deficit
  • Many variations of AI including natural language processing, machine learning, forecasting, pattern recognition and more.

Please read my blog on this for more depth on each.

Oliver Guy

Basic forms of AI, in the forms of sales forecasting, price elasticity, markdown optimization and other calculations, have been in use in retail for many years. History shows us that change management around these revolves around faith from the users — if they “buy into” the decisions and recommendations then things move much faster.

John McIndoe

Most retailers are proceeding cautiously with the testing and deployment advanced of AI along with machine learning (ML), as they should. Stephen Hawking summed up the power of AI with his comment, “Success in creating AI would be the biggest event in human history. Unfortunately, it might also be the last.”

The good news is there are high-quality AI/ML solutions available that are both powerful and easy to deploy. To maximize the impact of these solutions, retailers should ensure they are working with superior quality data sets and creating the ability to rapidly and effectively act on the insights the analytics generate. Effective preparation will go a long way to ensuring retailers are able to address AI-related questions.

"A lot more care and consideration needs to be put into introducing and rolling out all new technologies in the retail space."
"the more effective approach is to take a look at your prioritized business challenges and see which processes could improve via augmented capabilities."
"I’m one of the first people to embrace a new technology, but I gotta say – people need to be way more careful about AI than they are being."

Take Our Instant Poll

How important is explainability to achieving success with AI?

View Results

Loading ... Loading ...