Since the release of OpenAI’s Chat GPT last year, artificial intelligence (AI) has set off a modern-day gold rush in the business world — with companies across industries jumping on the bandwagon, eager to adopt AI and take advantage of the benefits that come with it. But the move into AI can be a risky one for businesses.
In a recent open letter, technologists publicly lamented the existential risks AI poses to humanity, and urged society to move cautiously into the AI era. While I’ll leave it to others to comment on those specific risks, we do see major pitfalls to companies rushing into AI too quickly — pitfalls that may present existential risks to their business if a crawl, walk, run approach isn’t adopted.
The promise and the pitfalls of AI for consumer brands
Artificial Intelligence systems excel at complex tasks that humans would struggle with, often by sorting through massive amounts of data (think petabytes!) and intelligently categorizing it. A perfect example is the way that Google photos can intelligently search your personal images, rapidly analyzing each pixel as a discrete data point, to identify all that include a “cat” just because you asked it to.
Consumer goods businesses — those that sell through retail and ecommerce — will realize enormous benefits from AI, and these benefits are not off in some distant future. They’re here today.
Every day our customer base of global consumer brands use our AI capabilities to monitor retail forecast accuracy, analyze unconstrained demand and use forward-looking simulations to predict where products will go out-of-stock before they happen. These types of capabilities are leading to millions of dollars in new revenue and cost savings, and major efficiency gains for the companies employing them. And while we’ll continue to build on this foundation of AI capabilities, we also believe that the industry should not look to AI just because of recent advancements and a lot of current hype. Brands should look to AI when it is useful and aligns with their business objectives. There are plenty of technologies we help our customers leverage that are huge value drivers, separate from AI — such as automation, data harmonization, statistical models and dashboarding.
AI must be a means to an end, and never the end itself. Before diving into AI, brands must start by asking themselves: What is the job to be done, and is AI the right tool to accomplish that?
If you want to be AI-ready, first, get your house in order
If we’re honest, consumer brands have been notoriously late adopters when it comes to modern technology and processes — especially with regard to data and supply chain.
Source: Geoffrey E. Moore, Crossing the Chasm
Go take a look at many of the most popular tools in the supply chain space, for example. It will be immediately apparent that most of today’s consumer goods supply chains are not built on modern technology. (Many of these platforms, in fact, hit the market 20 or 30 years ago). The platforms aren’t designed with usability in mind and extracting insights from data sets is a laborious, inefficient process.
Consumer companies have myriad data silos and antiquated processes that are not equipped to handle the scale of modern data in a responsive way. You only have to look at how many companies in 2023 are still:
- Manually pulling, harmonizing and analyzing data in Excel, a tool that is simply not built to handle data at the scale and speed that the market demands today.
- Struggling with data silos. When data is siloed between systems and teams, decisions are made more slowly, and are often based on conflicting or completely inaccurate information.
- Following a traditional S&OP process, spending a massive amount of human capital manually piecing together forecasts, massaging numbers and then spending weeks negotiating and reconciling the numbers, before they even begin to make decisions and adjustments for the future.
All of these inefficiencies introduce an amount of latency that is unacceptable in the modern business climate. If you calculate the bad business outcomes that result from these kinds of poor data practices (lost sales due to out-of-stocks, unproductive inventory, wasted marketing spend, just to name a few), the costs are staggering. Our analysis shows that the average consumer brand with $250M in revenue is leaving between $3.2M and $7.8M on the table annually.
That’s the bad news.
You wouldn’t use a sledgehammer to crack a nut
The good news is that none of these problems require complex AI models to solve. You wouldn’t use a sledgehammer to crack a nut, and likewise you shouldn’t use AI to solve problems when a simpler, faster, less costly solution is actually better. We have the technological infrastructure and data science expertise to support this level of computation. We could use AI to solve any number of pattern-identification and categorization problems, but that doesn’t mean we should. For our customers’ data (which is much, much, much smaller than the data filtering through Google’s image search engine), a simple linear regression model will often produce the same results as an artificial intelligence system at a much lower cost and effort to our customers.
Much of the data brands need to make better decisions is available today, but it’s not being utilized correctly. Take Walmart for example: Retail Link (Walmart’s data portal) gives you a hundred different attributes and metrics to understand consumer behavior, including sales, inventory levels and forecasts. But most consumer goods organizations haven’t invested in being able to extract insights from the data at the daily store/SKU/location level — a granularity necessary for tracking in-store execution (phantom inventory, on-shelf availability, out-of-stocks); measuring events, promos and item launches/discontinuations; allocation decision-making and more.
These data issues should be addressed before a brand fully adopts AI. After all, AI models are only as good as the data sets they’re trained on. An AI model is always learning; don’t teach it bad habits. If you have an AI model based on bad data, you’ll get bad decisions, recommendations and predictions back. So make sure that your data practices are in order first to protect yourself against these pitfalls. The kinds of AI models that help consumer brands require structured data, so get your data in order and establish practices to capture that data moving forward in a structured way.
Consider this step one on your AI journey.
Start with the low-hanging fruit
We work with many of our customers to reach a point where they have clean, accurate, harmonized data that isn’t siloed across teams and systems. At that point they’re a mature organization ready to step up to our more advanced AI capabilities.
But if you’re not there yet, you may want to slow down. Not addressing data silos and visibility challenges today is an existential risk to brands in the era of AI. The losers in this new world are going to see their market share evaporate, and they'll lose credibility with retailers — who expect brands to be data-driven in their recommendations. If you’re taking a month or longer to even understand consumer demand shifts, and only THEN making adjustments to your business, you’re already behind the 8-ball.
Adopting AI for consumer goods companies requires a “crawl, walk, run” approach, and most organizations try to go straight to “run.” But there is a ton of low-hanging fruit that can grow a company’s bottom line, just based on the data they have today, and you’ll want to ensure your focus on AI doesn’t come at the expense of higher-impact, short-term gains. AI is absolutely valuable for sophisticated, advanced organizations, but the vast majority of CPG companies have existing data and process problems that they’re not attacking head on. Tackle those first, and you’ll be poised to really take advantage of the benefits and avoid the pitfalls of the AI revolution.
