A look at where AI stands in the retail realm

A look at where AI stands in the retail realm

Artificial intelligence technology is on the cusp of huge adoption with global retail spend increasing to $7.3 billion annually by 2022 — quite a spike from the $2 billion spent in 2018, according to a Juniper Research study.

By 2024, AI investment is expected to reach $110 billion, according to an IDC forecast.

That spend will likely focus on boosting customer service and customer sentiment data as retailers strive to improve the customer experience by using AI insights on why and how products were bought and the level of service provided.

Juniper expects retailers will also employ AI for product design and development as well as crafting promotions and marketing.

"AI has a huge transformative potential for retailers, in a way that is difficult to underestimate, a potential based on changes underway in the retail industry, which are making AI more applicable all the time," Juniper noted in a whitepaper.

Yet, as reported by Forbes, most retailers may not see much return on the investment, at least in the initial efforts. A MIT Sloan Management Review and Boston Consulting Group found that just about 10% of 3,000 company leaders are seeing significant benefits from AI spend.

Retail Customer Experience reached out to Seth Earley, CEO of Earley Information Science, and author of "The AI-Powered Enterprise: Harness the Power of Ontologies to Make Your Business Smarter, Faster, and More Profitable," to get insight on where AI is today within retail, the challenges of adoption in the retail environment and what's to come in the near future.

Learn why there is no "magic" when it comes to AI and how information architecture is not only the starting point, it's the heart of AI.

Q. Where does AI "sit" today within the retail environment in terms of adoption and what will be needed to drive greater adoption?

A. I would say mostly in its infancy. Most retailers do not have deep knowledge of the AI algorithms (programs) that drive AI or the data that powers the AI programs. In some cases, they are experimenting with mostly self-contained tools that can perform some degree of personalization through more mature approaches like shopping basket analysis (where what you have put in your basket is compared with similar purchases by past customers) to make recommendations (people who bought this, have also bought this other thing).

But true personalization combines information about past orders with other, more valuable signals: demographic information, click path, real time and near real time responses to offers, email messages, marketing campaigns, interests derived from patterns of interactions and more. In the case of presenting 'bundles" with related products, compatible solutions, or recommended combinations, while various aspects can be automated and streamlined, much is left to the imagination of merchandizers and category managers. This is a very deep topic - many of the details of harvesting and acting on these personalization "signals" are covered in chapter three of my book — "customer experience — the front line of the battle."

One thing that is holding retailers back is the proliferation of tools in their marketing and customer engagement stacks and the disparate and sometimes incomplete or incompatible data provided by these systems. In many cases, additional tools are needed to consolidate customer interaction data (such as a customer data platform) and then there are multiple strategies to use that data to personalize or at least contextualize the experience.

More sophisticated approaches require extensive revamps to content strategy (including the use of componentized content to experiment with offering combinations), to product data structures, new treatments of the customer journey (using high fidelity journey models) and to ways that knowledge about customer needs is applied. Effective AI cannot be "bolted on" – it requires a great many supporting processes and solid information infrastructure as well as a harmonized marketing and customer engagement technology stack. In my book, I include some charts and checklists to help companies determine the maturity of their ecommerce systems, and to assist them with advancing to the next level.

Q. What are the most common mistakes retailers make when it comes to AI and are there any misconceptions?

A. Where do I start! That AI can be applied without deeper understanding of the customer, that AI is a silver bullet, that bad data is not an obstacle, that AI tools can fix your data, that one AI tool can do it all, that AI will replace entire functions, that personalization is about getting the right tool. There are many misconceptions. It's not magic. If building personalization functionality, someone has to determine what messages are appropriate for one segment versus another — unless that is, there is enough data for the too to experiment with to optimize messaging. But even in that situation, a human needs to build the personalization framework. AI set up needs human judgment. When "training" a helper bot or virtual assistant, the training data is knowledge and content — the same that you need to train a new employee. If organizations understood that, their knowledge, help and support systems would not be such a mess. Some organizations are creating new "AI content" functions. Guess what? That is "content" not just "AI content". AI can add a lot of value in many areas, but it begins with understanding end to end processes, value chains, information flows and customer journeys. You can't automate a mess and you can't automate what you don't understand.

Q. Why are key elements to successful implementation of AI (i.e. internal strategy)?

A. Retailers need to do the following:
• Identify a clear business objective.
• Mapping the end to end data, content, and knowledge flows.
• Measuring the process that the AI is to impact to get baselines.
• Defining a clear future state with metrics that will tell you if you are successful.
• Measure outcomes and compare to baselines.
• If you were not successful, deconstruct the process.
• If you are, look for ways to expand.

To sum up, having a clear understanding of the problem you want to solve and how you will measure success.

Q. AI produces big important data but parsing through that can be arduous. How can retailers help employees being crushed by "information overload"?

A. Information overload is a misnomer — it really is filter failure. We can be overloaded with extraneous information no matter what tool, system, process or location. The key is to filter out the noise and focus on the objective and relevant information. This starts with mapping the internal customer journey — what I call the "high-fidelity journey map" in my book. What are your customers trying to accomplish, where do they go and what information do they need? How do they think about that information? What is their mental model? Test their use cases and measure baselines. Then take a narrow slice of the day in the life and focus on an important, measurable work process. The information that supports that process has to be structured, curated and segmented from the extraneous information. That is usually done with a content model and a method to access the content. It is personalization and contextualization on an internal basis. How can we get the right information into the right person's hands at the right time? The only way to do that is to understand that task and what they need and then throw away (or archive or segment) the unneeded distracting information. There is no free lunch. Many are trying to use AI to make up for their past sins in poor data hygiene. It's catching up and while artificial intelligence can help, it begins with the blocking and tackling or use cases, scenarios and content strategy. There's no magic. It begins with the information architecture (the so called "IA"). There is no AI without IA.