Photo: Getty Images/PeopleImages
Are retailers set up to scale the value of AI investments?
According to McKinsey, the retail industry has the potential to create $1.7 trillion of value, or 12.39 percent of total sales, from artificial intelligence.
How much value do you think AI creates for a retail business?
Before answering such a question, you must first have established goals and the capacity to track use case performance, plus a way to translate it into metrics that matter: sales, margin, customer value, retention, productivity or efficiency.
This is important because companies that achieve outsized benefits from AI fund their use case roadmaps on the backs of successful projects. You simply cannot do this if you approach AI within silos and buried within business processes.
AI is most often employed to automate tasks within processes. This means you have to isolate the impact of AI on the process in order to measure effectiveness.
Fortunately, for those early in their AI journeys, most initial use cases that pass muster during prioritization are based on lots of historical data just ripe for machine learning. So, the baselines for use cases targeting improved personalization, recommendation or recognition should be readily available.
Retailers already proficient with AI have built up the technical, organizational and governance foundations necessary for significant value scale. Your goal should be the same, but you cannot get there overnight.
That is why before you head down any AI use case path you should:
- Ensure you have the means to establish a baseline from which you expect to improve.
- Have the data to support the baseline analysis — data that is updated at the right frequency and is available in the right format; data that connects actions performed by AI to business metrics.
- Be capable of conveying use case results in business metric terms to the executives who fund and depend on the AI.
- Plan for next steps: If results fail to meet expectations, make changes or retire the use case and leverage the learnings into the next one. If results are positive and your executives seek to improve further, build upon what you have.
- How to set expectations for your AI project – LinkedIn/ Gib Bassett
- The Executive’s AI Playbook – McKinsey
- How to Choose Your First AI Project – Harvard Business Review
- AI, ML and Deep Learning – A very quick primer for curious executives – LinkedIn/Rama Ramakrishnan
- What are the biggest barriers to AI adoption for retailers? – RetailWire
BrainTrust
Rodger Buyvoets
CEO/Founder, Crobox
Cynthia Holcomb
Founder | CEO, Female Brain Ai & Prefeye - Preference Science Technologies Inc.
Chuck Ehredt
CEO, Currency Alliance
Discussion Questions
DISCUSSION QUESTIONS: How aware do you think retail executives are today of the impact AI is having or could have on their business? What do you see as the main barriers to retailers tracking the performance of AI use cases?
Execs may note the importance of AI but not understand how best to use it. And yet it’s true that leveraging AI should be a mandate from the C-level down. Without a culture of experimentation at the core of a brand’s strategy, it’ll be more difficult to reap the benefits of AI. In order for this to stick (and work), retailers need to be data-driven, flexible, and IT-first – often a hard sell in an industry dominated by legacy issues and interdepartmental silos.
Well said Rodger, I agree!
They are probably aware and probably unsure as to how it is best used. They need to differentiate it from automation, which is where productivity gains might be found. Hard to believe there’s trillions in value being left on the table – sales are a relatively zero-sum game so if somebody is picking up value with AI, somebody is losing. And that is the biggest obstacle – unrealistic expectations about what AI can do for you short term and long term.
I also thought about the zero sum thing Dr. Needel, there are definitely winners at the expense of others when it comes to this value.
Often, when there is a new technology solution in the market, there’s a lot of media attention centered around the theory of technology instead of its practical use. AI can do many things, but it’s only helpful to retailers when used to solve an existing priority business problem faster, better, and cheaper. For example, AI is great for personalized product recommendations. Amazon generates 35 percent of its sales through its recommendations engine. Similarly, Halla’s AI engine is trained using recipe data and natural languages to easily plug into a retailer’s e-commerce and/or scan and go platforms.
Agree about product recommendations, probably the #1 most talked about AI use case in retail, but also the least understood when it comes to what makes a great one, versus a less great one.
Retailers are not aware of the potential impact because few use cases exist that show that AI adds tremendous value. It’s challenging to measure the impact and costly to implement solutions, especially data-focused solutions because the systems are not in place to enable it. Outside the standard e-commerce recommendations, demand forecasting, intelligent routing scenarios, and some robotics, there haven’t been major strides in applying AI to retail. The big drawbacks are tied to lack of data infrastructure, real-time data access and AI/data talent in the retail space. Regardless of retail exec awareness, retailers are unable to move towards major AI powered results without a data science foundation. For retailers it’s a matter of practicality.
Thanks Ananda. I think retailers are in a tough spot. On one hand as you say, there are many AI use cases used today but the value may not be the breakthrough execs expect or desire. The problem is that these use cases, IMO, were done in relative isolation and not part of a thoughtful plan for AI across the retail organization.
There is some good research from Tom Davenport showing that the transformation value from AI for most companies in any industry, at their initial stages, comes from adding up the value of many use cases and leveraging these into successively more valuable and potentially more complex use cases that have massive value potential. If execs don’t recognize and believe this, status quo results and unfortunately probably poor financial results too.
AI deserves the attention it is getting but, for the algorithms to work, they need clean data – and too few organizations are dedicating the effort to clean up their data and generating single customer profiles. Without those foundations in place, a great deal of money will be wasted on AI as the ROI will be depressed by poor data management practices.
Thanks Chuck – agree about data quality and foundation as important. What I have seen is too much of this data quality conversation confined in a small group, usually IT, without connection or direction from the business. This persists and over time unfortunately I am sure it hurts the company’s performance and it struggles on many fronts with fewer resources due to COVID and general market headwinds.
In many businesses, AI is still a solution looking for a problem, so its potential to drive improved business outcomes is not well understood. Analytics teams need a C-suite sponsor behind a strategic initiative that puts AI’s potential to work in a visible way. They will also need to take the black box off of AI to help explain the “why” behind the outcomes it drives in a way that can be broadly understood and ultimately increase its leverage. Only then will AI be recognized for the impact it is having and be fully embraced as a meaningful business tool.
Well said Kim. It would be great if there were an Executive Accelerator to get over the hurdle of understanding and trusting the potential, such that an Exec had the confidence to stick their head out a bit into an unfamiliar area.
Retailers are undergoing rapid omnichannel transformation and AI can be a key weapon. Given the need for speed, retailers should prioritize use cases and leverage proven, ready-to-go SaaS solutions wherever possible. Internal data science teams should focus on high impact use cases not readily available.
In our current environment, AI speed will eat the competition.
Thanks Craig, your wording here is exactly as I have described it as well — balancing what you can do fast that has value via packaged means, while exploring cases that are maybe more complex and specific to your business such that you may need data science skills. The challenge I’ve observed is that use cases across these two scenarios tend to live in silos so there is no reconciliation of business value relative to priorities, resources and budgets. I think those that do however stand to get a chunk of that potential AI value reported by McKinsey.
Retailers and all who engage and touch internal AI initiatives need to understand the basic premise of AI and machine learning.
Today, leveraging AI as a business intelligence solution relies upon a programmer or an engineer to unilaterally decide what is the “correct” answer to the AI-enabled solution. Unfortunately, data is not smart, only words strung together, waiting for meaning to be attached to them by a human who is designing an algorithm. Thus enter the data scientists and the programmers, hired to find a solution to a business problem they have never experienced in the real world. Their mission: curate the correct data points out of huge data silos to solve for and validate the answer to the prediction the programmer/engineer/data scientist has determined is the “correct” answer to the problem they are tasked to solve by the retailer.
Retail leaders must get into the guts of the problem they are asking data scientists and programmers to solve. Only then will they be able to determine the ROI of a successful AI implementation. If retailers and their teams become intimately engaged in the “deep soup” of data, they will begin to learn and determine for themselves what is the “correct” answer to the AI initiatives they are funding.
Data is not magic. AI is not magic. The magic begins with experienced retailers investing themselves into learning, getting their hands dirty, able to determine for themselves what is the “correct” answer. Awaiting the retailers who dare to engage? Immense opportunities of AI enabled, proprietary solutions reflective of the retailer’s customer, brand and ROI goals, leveraged by their secret weapon — retail expertise.
Very thoughtful comments Cynthia. You describe the bespoke “craft” data science-led use case path that few inexperienced leaders in any industry have time or the inclination to risk exploring. It’s unfortunate given the potential opportunity and as you say, bringing the industry and business knowledge heat to solving the right problems can pay off. The flip side are the use case opportunities to accelerate AI projects within packaged software, which I happen to be familiar with. The best ones allow the retailer to tailor AI to their specific needs, and done right, allow the retailer and the associated business, technical and analytical teams to really learn in practice how AI use cases become reality and are different from how a process unfolds today – the skills, data, ethical considerations, governance issues, and technical architecture and integration requirements. Risk is lower, time to value faster, but if you don’t manage these use cases as a group IMO and have a thoughtful strategy to expand upon win after win, then the retailer may never realize any significant value from AI investments, be they bespoke data science fueled or otherwise.
Aside from skills, arguably the biggest barrier to harnessing the value of AI is data. Data in retailers is notorious for being siloed and difficult to extract, combine and exploit. It is all very well C-level saying “let’s exploit AI” but if the building blocks are not accessible then it will not work.
The next stage is actually being able to execute the result. As an example, one of the biggest areas of benefit is price and promotions, coming up with a AI determined promotion or price change is likely to be let down by how long it takes to be executed in the real-world. The lag between a head-office decision and the price/promotion changing in store is often measured in days — contrast this to Amazon who routinely change prices multiple times a day balancing demand, supply and price elasticity.
Retailers are yet to invest in real-time store IoT platforms that would allow rapid execution of optimized decisions. This would mean that promotions could be modified when products were selling too fast or too slow — thus maximising overall margin.
Very good comments Oliver, I totally agree with the value of the use cases you cite here and the necessity of the real or right time technical requirements. I can’t remember exactly but I think a company like Kroger perhaps with the support of 8451 does the things you describe. Even if I’m not recalling this example correctly, the solution scope and expertise necessary unless available truly turnkey and highly packaged, is arguably hard to justify as the initial use cases in a retailer’s AI roadmap. The research and best practices I’m aware of suggest that the value of AI use cases must accrue and be recognized by executives as a means to fund ever higher value and possibly risker but more valuable use cases. I think this makes complete sense but the way most retailers think about data, IT, management and processes I fear flies in the face of what’s needed to really achieve scale value of AI investments. I think most retailers interested in real time IoT enabled adjustments in price and promotion at the shelf must test what this buys them versus the current state to justify the project. To get there, a small market test done more manually, but leveraging ML to derive or leverage unrecognized factors, might prove the utility of a broader and real time architecture investment.