Retailers: Is AI the answer to all your problems?

Justin Fischgrund

Justin Fischgrund is the Founder & Principal of Fischgrund Consulting, a niche firm specializing in process improvement and technology strategy for organizations across industries. Justin is an engineering and technology leader, patent inventor and professor, with a proven track record of pioneering process improvement and cutting-edge, technology-driven solutions across diverse industries, including financial, naval, aviation, railroad, retail, and grocery.

Over the last few months, the most common questions I hear from clients surround one topic: artificial intelligence (AI). Whether it is “how can I adopt it?” or “will it replace my job?”, the pervasive curiosity surrounding AI remains undeniable. 

From colleagues at work, to my own grandmother and my one-year-old son (now, that’s a stretch), there’s a universal desire to delve deeper into the realm of AI. But don’t just take my word for it; let’s look at the stats. According to a recent study, grocers plan to increase their AI spending by 400% before 2025. The same study found 73% of grocery tech executives expect AI to be embedded into most or all of their software capabilities by 2025.

So what should matter to grocers when it comes to AI? Read on.

The basics

So, let’s dive in. Starting with the fundamentals, AI simply refers to the simulation of human intelligence in machines, which enables these machines to perform tasks that would typically require human cognition. Within the umbrella of AI, various types exist, each with distinct applications:

 Although AI has been a fixture in various industries for years, the emergence of GenAI has introduced the ability to create brand-new content based on existing data – growing the potential reach to unprecedented heights. Whether it can enable your organization to optimize inventory allocation, augment customer service interactions, define dynamic pricing models or enable personalization within a shopping experience, AI has a place in any organization. So, you might ask, “Why shouldn’t I be adding this to our roadmap to kick-off immediately?” 

Depends on the data

Let’s pump the brakes for a moment. Don’t misunderstand me; AI has the potential to transform your organization and redefine your future. I’m not here to dispute that fact. However, implementing AI isn’t as straightforward as saying, “We are going to start using AI to drive operational efficiencies.” Successful implementation hinges on multiple factors, most notably around the data and the why. Firstly, your business needs to be able to answer the most important question of all: What problem are you trying to solve? This clarity of purpose lays the foundation for success. Without a clearly defined target, any solution will lack direction.

In addition, the efficacy of AI systems is contingent upon the quality of the data they rely on. Therefore, cultivating a robust data infrastructure with transparent decision-making becomes paramount. In essence, if you lack sufficient data or the ability to comprehend and scrutinize the model’s decision, your AI initiative becomes futile. Imagine attempting to allocate inventory across a region of stores without adequate sales or customer data, or only having data through 2022. Pretty difficult, right? Now, imagine a similar dilemma when trying to devise an updated clearance pricing strategy. Envision receiving insightful analysis from a colleague, but when questioning the specifics, the underlying reasoning can’t be revealed. Ready to present this to the CEO? Complete and transparent data are crucial for any analysis. This becomes even more essential when dealing with AI-driven analysis.

Equally crucial is ensuring your organization possesses the requisite resources and personnel structure to fully leverage AI’s potential – but I’ll spare you my soapbox speech on that. Just remember, without the right team, even the most sophisticated AI initiatives can falter. 

“Garbage in, garbage out”

Now that I’ve delivered the bad news, let me pile on some more. Beyond the primary considerations of value, data quality and team structure, it’s critical to acknowledge and mitigate various risks associated with AI implementation. While these risks aren’t insurmountable obstacles, they demand thorough comprehension and proactive mitigation efforts. 

Most importantly, AI is just a computer – it lacks human intuition and common sense. For example, if you want AI to tell you the “best” way to do something, good luck. On top of this, I’ll say it again, AI is only as good as the data that feeds it. Have you ever heard the term “garbage in, garbage out?” That’s a challenge when you have a human deciphering the data. Imagine the exponential risk introduced with computer decisioning. 

See for yourself

Beyond data quality, AI may harbor bias if sources favor one viewpoint, resulting in outputs that are myopic and predisposed, potentially inaccurate and unrepresentative of reality. Also, many AI models rely heavily on the prompt that feeds it. Clarity, contextual clues, open-ended questions, and patience are necessary components when dealing with prompting AI. I wouldn’t be surprised if “Introduction to AI Prompting 101” is a college course in a few years. If you don’t believe me, try this experiment: Input the below prompts into any GenAI tool & compare the outcomes. Notice any differences?

Artificial intelligence presents vast opportunities for organizations to optimize processes, shape strategic priorities and achieve financial excellence, especially during unique macroeconomic conditions. However, having a deep understanding of its limitations and associated risks is critical to ensure responsible and effective implementation. 

Integrating ethical analysis into design and deployment phases, mitigating biases, promoting transparency and ensuring organizational readiness are vital for responsible AI adoption. If you can lay the foundation and establish the proper controls, AI can revolutionize your business. But until then, proceed with caution.