How data literacy allows gen AI to drive productivity at Dow

With a huge amount of innovation going into the gen AI products marketplace, large, global companies must improve their ability to leverage these technologies. But what does it take to build a culture that embraces gen AI? For Melanie Kalmar, the answer is data literacy and a strong foundation in tech. Learn how she, her team, and the executive committee create a “gen AI ready” culture.

How do data and digital technologies impact your business strategy?

At the core, digital at Dow is about changing how we work, which includes how we interact with systems, data, and each other to be more productive and to grow. Data is at the heart of everything we do today, from AI to machine learning or generative AI. This work is not new to Dow. We’ve been leveraging predictive technologies, or what I call traditional AI, across our enterprise for nearly two decades with R&D and manufacturing, for example, all partnering with IT. 

What are a few examples of these traditional AI capabilities?

Let’s take the reliability of the furnaces we use to crack ethane into ethylene. We’ve built digital twins for several furnaces we operate across the globe, and we currently have 70 AI models running on those furnaces. These models allow us to predict failures early, and we forecast a 20% reduction in furnace unplanned events, improving repair times by at least two days.

So AI helps us have fewer emergencies. We’ll still have to decoke, or clean out, the furnace, but now we can plan for it. We can offload the work to other furnaces and avoid emergency mode. These huge productivity savings also drive sustainability improvement by allowing us to minimize energy consumption.

We’ve also leveraged AI in the supply chain to revolutionize our demand forecasting and supply network planning. We’re now able to provide real-time predictions about our network performance, optimize our inventory, and reduce costs.

I’m a firm believer that the planning function is at the core of getting the right products to our customers, because we’re producing them at the right time and in the right places. As an integrated manufacturing capability, Dow is a complex puzzle, and these AI models help us incorporate historical data, market trends, and customer behaviors, all of which allow us to produce a more precise demand plan. Ultimately, our use of AI is all about being a reliable supplier to our customers, and it’s working: last year, we had our highest ever customer experience index metrics.

Is gen AI a shiny toy in a hype cycle or will it have material impact on your business?

We’re implementing Microsoft 365 Copilot and seeing amazing results already. We surveyed our early pilot users regularly, and when we asked them how much time they were saving daily, we thought we hadn’t worded the question correctly, because more than half told us they’re saving one to two hours a day. We’re now expanding the user group from a small subset to a third of our employee base, most of whom are in an office environment. Several groups are already recognizing cost saving opportunities alongside efficiency gains. For example, in public affairs, our colleagues are using gen AI for first drafts of content development, or to analyze vast amounts of data. It’s helping identify emerging trends, public sentiment, and potential issues so the team can proactively address challenges and seize opportunities.

What are additional examples of early signs of productivity from gen AI?

Our teams can ask Copilot to “review my email and prioritize what I should focus on today,” or to “find that PowerPoint I was working on last week.”  A significant Copilot use case has been finding documents.

Patents are another key area for gen AI. With patented solutions, we have opportunity for higher margins. So our R&D teams are always looking for the next molecule or a new way to solve a customer problem. Traditionally, they’ve had to sieve through reams of patents to determine if their current work is patentable. But because gen AI can assess thousands of relevant patents in a short period of time, R&D has reduced this patent research from four months to four hours, in some cases.

What was the foundation you needed build to benefit from gen AI?

From a technical perspective, we needed a modern, secure network. For years, we’ve been partnering with companies, including Cisco, for our network capabilities, and we’re leveraging the hyper scalers. We also have a blended architecture of deep process capabilities in our SAP system and decision-making capabilities in our Microsoft tools, and a great base of information in our integrated data hub, or data lake, which is all Microsoft-based. That’s what we’re running our AI and our machine learning against.

But the technical foundation is just one piece. For gen AI to drive real productivity, we had to build a culture that adopts new solutions and appreciates the value of digital capabilities to streamline their work. AI is not about layering tools into the organization. It’s about helping people to work smarter and be more productive. Before we started rolling out AI capabilities, we had to improve data literacy.

How did you do that?

Dow is implementing data literacy as a key component of our digital transformation strategy. The vision is to empower everyone to speak the language of data and equip them for success in the digital world. This involves changing the collective mindset, language, and skills to be open, willing, and curious about data, and to apply data constructively in solving problems.

Our data and analytics literacy program includes a curriculum of on-demand content developed using the expertise of internal professionals to generate Dow-specific content, as well as content delivered via Coursera, a leading online learning platform. We’ve had great success within our own IT function leveraging Coursera learning, recently surpassing a 92% participation rate on AI literacy. We’ve also had great C-suite participation. Later this year, we’re launching an enterprise-wide data literacy program, too. This initiative is part of Dow’s broader activities to make AI scalable and sustainable within the company.

How have you broken up the topic of AI into a literacy program?

I believe that seeing is believing; don’t just talk to me, show me something. For the board, we recently conducted a gallery walk highlighting AI. We had four small teams demonstrate concrete examples of the AI capabilities we’re working on. The board interacted with solutions and asked questions, which helped get us all on the same page about how we’re using these new tools. Also, last August, we ran an AI immersion day, which the CEO Jim Fitterling and I co-hosted for our top 200 leaders. The day was about exploring the potential of generative AI and what the benefit could be for our customers and employees. We brought in some external speakers, and we spent time brainstorming in table groups about how we can leverage AI.

We started by explaining the difference between traditional AI, which we’ve been using for years, and generative AI, which is new. We also shared videos of Microsoft 365 Copilot and we jumped into a discussion of other gen AI-based companies. The point was to open minds, and to demonstrate we want the whole enterprise leadership team to impact what we’re going to work on. They got very excited about having a role to help their teams leverage AI to become more productive. We came out of that session with over 200 ideas, which we bubbled up into 20 categories. Then the leadership team voted on the top five ideas with the highest value-generation potential for the company. I’ve also asked the executive suite and my entire organization to take some gen AI fundamentals courses online.

How are you leveraging data scientists at Dow?

We operate a hub and spoke model that has a centralized IT group over governance, and data scientists across the company in manufacturing, supply chain, and R&D. Our data scientists can’t only know data science; they need to understand the core of our businesses so they can use data to solve business problems.

Several years ago, I organized a Dow Data Science challenge, where the former CFO and I brought together our data scientists and presented to them a big audacious challenge: Add $100 million of EBITDA to the company. We put them on teams with people they don’t typically work with to engender new ideas. We had so many actionable ideas from that challenge that we’ve done it every year since. When you bring together data scientists from different functions, that’s where the innovation and the enterprise solutions happen.