Welcome back to our “Customer Service Heroes” series, where we invite inspiring customer service leaders to share their advice for running successful teams.
Carolina Pinart is the AI Program Lead at Nestlé, the world’s largest food and beverage company. Nestlé operates more than 2,000 brands across the globe spanning bottled water, petcare, chocolate and confectionary products, baby food, cereals, and more. In her role, Pinart is responsible for integrating emerging technologies into Nestlé’s products and business operations.
Many people look at Nestlé and strictly see a consumer packaged goods company. In reality, though we’re not a traditional technology company, there are so many technology tools that are required to operate a company of our size on a daily basis. We sell one billion products every day and rely on a huge global supply chain infrastructure. As such, all of us at Nestlé are constantly trying to innovate how we use technology to optimize and automate our manufacturing processes, marketing materials, research and development techniques, and more. To be a global leader today, every company must be somewhat of a technology company.
Nestlé’s journey with AI
As with most other CPG companies, Nestlé began as a very product-centric business. Over time, we’ve evolved to be more customer-centric. We don’t just want to provide dog food to a consumer, for example. We want to understand why that consumer purchases a certain dog food, how their pet fits into their life, and what motivates them to make a purchase. With that data in hand, we can get smarter about cross-selling other products and providing more personalized recommendations to each customer.
In order to become a customer-centric organization in this way, Nestlé must rely on artificial intelligence solutions that allow our team to analyze large amounts of data, communicate directly with customers at scale, and create comprehensive customer profiles and personalized campaigns. Those communications and campaigns could be in the form of chatbot interactions, interactive features on our website, or personalized recommendations sent to a customer via email.
In addition to using technology to aid our transition to a customer-centric organization, we also need to use artificial intelligence to inform our product development and manufacturing processes. Artificial intelligence can have huge implications for how we optimize and automate our manufacturing and supply chain operations. A “simple” 2% improvement equates to billions of dollars in cost savings when you’re talking about an organization as large as Nestlé.
That’s where I come in. There are so many questions — on the customer-facing end and the business operations side — about how emerging technologies like artificial intelligence can help us. And typically there are no plug-and-play solutions. Each piece of technology needs to be customized to suit our needs and dependencies need to be built to satisfy our internal processes. With a chatbot, for example, we need to determine how a bot would work alongside our human customer service agents, what topics it would manage, what would happen if the bot broke or malfunctioned mid-conversation, and how we’d evolve the use of our bot over time. My team is responsible for determining answers to all of those questions so we can make technology tools work for all parts of Nestlé’s business.
Ethics in AI
One of the most important aspects of my job is considering the implications of using AI in our daily operations. The benefits and opportunities available with machine learning are huge, but machines aren’t perfect. At its core, machine learning is based on data. But where does that data come from?
Data used to teach machines is typically created or influenced by humans. And humans are naturally biased creatures. We hold inherent biases based on where we are from, how we were raised, and different demographic characteristics. Artificial intelligence that is based on human data will learn and inherit these biases. AI essentially acts as a magnifying lens because it takes data and amplifies it to the extreme — when that happens, we don’t want it magnifying the wrong things.
A prime example of AI bias gone awry is Twitter’s image cropping software that prioritized images of white men over black men or Amazon’s AI recruiting tool that unintentionally developed and amplified a bias toward female applicants. Even more ethical questions come into play when you consider the use of AI algorithms and tools for direct customer interactions. Facebook recently came under fire for using AI algorithms that prioritize engagement with content above all else — which can lead to the amplification of hate speech or content that can be mentally distressing.
At Nestlé, my team is working on solutions to integrate more ethical procedures into AI tools. It all starts with educating employees about unconscious bias and teaching employees to recognize bias in data sets that are later fed into machine learning tools. It’s a herculean task, but essential for our company because any time you have a machine that makes a decision on behalf of a person, there is a risk of discrimination or bias in AI.
Using human intelligence to combat bias in AI
One of the ways we can identify and solve for ethical concerns and bias in AI is by using something called “human-in-the-loop AI.” This is, essentially, the process of having a human being place checks and balances on AI tools to ensure they operate without bias (assuming you’ve been able to train bias out of the humans, as well).
In marketing or customer support communications, for example, humans can work alongside AI algorithms to identify cultural nuances in language translations. Not every word or phrase translates directly from language to language — humans can help machines account for local etiquette, figures of speech, metaphors, and more. The stronger the diversity of your workforce and local experts, the stronger your language ethics will become.
For Nestlé, this is similar to how we approach global product development. We have a research and development structure, but always take local nuance and insight from locals into account. Just think about our Kit Kat bar — chocolate is the go-to Kit Kat flavor in the US, but in countries like Japan there are hundreds of different Kit Kat flavors available. Flavors range from soy sauce to green tea, red bean, strawberry, and cherry blossom — all based on local market preferences.
We need to approach the sensitivity of our technology integrations in the same way we approach global product development. AI has the power to change the way we operate for the better — but it’s up to humans to teach machines the right way to behave.