‘AI Is a Business Challenge, Not an IT One; Data Is the IT Problem,’ Q&A with Armstrong World Industries

For Armstrong World Industries, AI strategy started with a practical question from the CEO: How is the company going to use AI?

The answer from the IT and data team was not to start with a tool. It was to step back and ask what kind of data foundation the business would need to scale AI across the enterprise.

Brent Lewis, Armstrong’s Senior Manager, Data & AI, discussed that work during a 2026 SAP Sapphire session on building a data foundation ready for AI. In a conversation with ERP Today, Lewis described how Armstrong’s history, M&A strategy, SAP landscape, and business need for faster answers shaped its approach to data and AI.

Armstrong is a 166-year-old company that manufactures ceiling, wall, and specialty architectural products. Its data environment has become more complex through business splits, divestitures, acquisitions, and integration across SAP and non-SAP systems. That complexity made AI a business opportunity, but also exposed the importance of data structure, governance, and scale.

(This interview has been edited for length and clarity.)

Q: You had a session at Sapphire. Can you give an overview of what you presented?

BL: The presentation was about our plan as an organization to have a data foundation that was ready for AI. We’re not a young company, and there’s been a lot of change over that period of time.

Our data has become very complex. We had a company that included flooring, ceilings, and walls. That split into two companies, and I represent the ceilings and walls company. That was a big data split. Then we divested our European and Pacific Rim markets, which created additional data complexity.

Our growth strategy has also involved M&A. We do two to three acquisitions a year, and that creates another layer of complexity, especially because those companies often aren’t on SAP systems when we bring them in.

We were having a hard time keeping up with the business need for good data. What happens is the business needs answers, so people export data and get an answer. That may work in the moment, but it becomes very difficult to control. Different people can end up with different answers.

Then AI hit the mainstream consciousness, and we realized quickly that AI is a business challenge, not an IT challenge. Data is the IT problem. We knew we needed a very different data foundation. For any kind of AI, that semantic layer is very important.

We structured our base data platform around SAP Business Data Cloud and Databricks, with other capabilities around that. The Sapphire presentation was about how we got there, why we did it, where we are today, and what our future ambitions are.

Q: How is AI a business problem and data an IT problem?

BL: Early in the journey, we saw consistently across research from firms like Gartner, KPMG, PwC, and others that AI cannot be treated as an IT fix. IT can help the business, just like with any application, but adoption won’t happen if AI is only driven by IT.

If AI is done right, it should accelerate the business strategy. Anything we do in AI has to tie back to the business strategy somewhere.

That means the business process being impacted matters. The business has to bring the value case, and IT helps support it.

I like an analogy I heard about electricity. When people first saw a light bulb, they thought that was electricity. But it wasn’t. It was a light bulb, which was a use of electricity.

AI is similar. AI is not the thing itself. It’s an integrated system behind the thing. ChatGPT, for example, is like the light bulb. The AI foundation behind it is the electricity.

We try to tell the business, ‘You tell us what your light bulb looks like, and we will make sure electricity gets to the lamp. We will help productionize it and make sure you can turn on 100 of those if you want.’

Q: How are you approaching the journey from data foundation to scaled AI?

BL: We use a crawl-walk-run approach. To scale AI successfully, we have to be able to run. If the vision is a fully autonomous organization, that’s running at full speed. It’s sprinting. If you try to go from nothing to sprinting, you’re going to fall and hurt yourself.

So we ask, ‘What does crawling look like? How do we get off the floor and start moving?’ It’s like watching a child learn to walk. You fall down, get up, and learn a lot of lessons along the way.

Now we’re at the walking stage. Our foundation is there, and we have two things in production that we can scale. We also have a list of things we plan to do this year. That’s how we start walking faster, with the goal of running by 2028.

Q: Do you have an example of an AI use case that has worked well?

BL: Pricing. We have a core business, which includes products like standard ceiling tiles. That’s about 90% of our business. We also have an architectural specialties business, which includes more bespoke products, such as wall coverings or column covers.

In that architectural specialties business, pricing works differently because of how the products are manufactured. Previously, someone would take data from our BW system and other sources, put it into a spreadsheet, and use that spreadsheet to provide budget pricing. We’re careful to call it budget pricing, not a quote, because it’s the first step in the process.

That person moved into another role, and the manager asked whether this could be a good use case for AI. We automated the entire process.

Now, a salesperson can interact with the system through a chat interface and ask for pricing on a certain type of product. The system asks the needed follow-up questions, uses the relevant data, and provides budget pricing. We also monitor accuracy and drift.

It helped in two ways. The person who had been doing the manual work could move into a more effective role, and the sales team could get an answer on the spot. It improved productivity and user experience.

It also improved visibility. Before, a spreadsheet would produce answers that were shared by email, and visibility could get lost. Now, the process captures data in real time and can feed dashboards that show pricing movement and buying behavior.

Q: Was that SAP-built or custom?

BL: It was custom-built. Business Data Cloud and Databricks give us flexibility. As an on-premise customer, we’re challenged in terms of how much Joule functionality we can access. But the connection to Databricks in our Business Data Cloud environment is robust.

We can connect to the data we need, share it through Delta, and serve that out for AI on the other side. I think many customers don’t realize how self-serving these tools can be now if they have the right team in place.

We also changed our team as part of the data transformation. You cannot change your process without changing your people. Now we have a team that can support Databricks, work with the business on use cases, design pilots, push them into production, and support the operating model around that.

Q: How is Armstrong’s relationship with SAP, and how does it shape the path to cloud and AI?

BL: We have been an SAP customer for about 27 years. SAP is a very good ERP system, and one of its strengths is how customizable it is. If you think about how a business differentiates itself from competitors, a lot of that shows up in customizations. The ERP system has to support that, and SAP has done that well.

The challenge for a customer like us is moving from a heavily customized on-premise environment into subscription models, cloud models, and now consumption models. That evolution makes the relationship more complex. We’re a publicly traded company and are sensitive to operating expense, so moving from capital expense to operating expense doesn’t always fit easily with our business model.

Our SAP system has been customized over many years to support how we differentiate the business. That customization now creates some of the pain in getting to the next model.

SAP and customers like us often want the same future state. For example, SAP would like customers to keep their semantic value in the SAP ecosystem when they build agents, and I agree with that vision. But as an on-premise customer, the question is: How does SAP help us get there?

The cost is high, and the business impact is high. We’re aligned on the vision, but getting from the old environment to that future state is the hard part.

Q: What are your customers asking Armstrong for, and how does that shape your technology needs?

BL: Our customers largely buy through a distribution system. We work through distributors to sell to end customers, so our customer interactions often go through those distributors.

As we have grown through M&A, we have added adjacent products and capabilities. Many of those acquired companies have their own brands and systems. That can make it difficult for customers to know where to buy a product.

About two years ago, we did a large effort to understand customer pain points. The biggest issue we heard was that it could be difficult to know where to buy. Customers wanted to buy from us, but they didn’t always know where to go.

We’re very intentionally trying to make the customer experience feel like one Armstrong experience, where customers know where to go and how to get what they need. That’s also a data and systems challenge because acquisitions often come with non-SAP systems that need to be integrated over time.

Q: How long does it take to bring acquired companies into the ERP roadmap?

If we stopped buying companies today, it would probably take about three years to get all companies onto the roadmap.

At two to three acquisitions a year, you need a strong plan to bring each company into the environment. We weren’t as intentional about that when we first started the M&A journey, so now we are catching up.

The timeline also depends on the company being acquired and the agreement structure. We don’t immediately integrate every company fully into Armstrong. We often integrate infrastructure first, and then there’s a plan for everything else, especially ERP.

Q: What advice would you give to companies that are growing through acquisitions and trying to manage data complexity?

BL: It depends on the maturity of the company being acquired. Do they have a strong system already, or are they working mostly out of less mature tools? That changes the approach.

One lesson for us was the importance of runbooks. We developed runbooks early for infrastructure, which is the first thing we integrate. You absorb the infrastructure, secure it, and standardize it. That’s very important.

Now we’re extending that thinking to the rest of the business transformation. As companies are acquired, what are the steps? What needs to move, why does it need to move, and when should it move?

My advice is to get that plan early, especially if you plan to acquire more than one company. The AI tools available now can also help. They can analyze data complexity, identify what needs to be transformed, and help create a more solid plan than what would have been possible a few years ago.

Q: Were there cost lessons from acquisitions that surprised you?

BL: Yes. When companies make acquisitions, they often plan for SG&A [Selling, General, and Administrative] savings and operational synergies. But that’s not always simple. Because integration is a roadmap journey, overlapping costs can remain for a while. You may get the synergy at the end, but in the middle, there can be additional cost.

Early in our journey, I don’t think we were fully mapping out the impact of bringing users and data into our systems. If an acquired company has 300 users on one system and we move those users into our Microsoft and SAP environments, what’s the licensing impact? What’s the sizing impact for data? Those weren’t always things we thought through carefully enough.

Now, when we look at portfolio cost, we filter everything through an M&A lens. We identify how much cost is tied to M&A so the business can understand the real impact.

Q: What are you focused on for the rest of 2026?

BL: From a data perspective, our 2026 goal is to put six use cases into production by the end of the year. We have one, so we need to get the other five.

That’s the walking stage for us, showing the business that we can do this. We put the plan in place, and now we can execute it.

The team is in place, the business is excited, and people are starting to understand what we can build. It has been really fun to watch that evolution.