What Supply Chain Leaders Get Wrong About AI: A Conversation with Microsoft MVP Dag Calafell

In This Article

MCA Connect’s Dag Calafell, Director of Technology Innovation, was recently named a Microsoft Most Valuable Professional (MVP) in the AI ERP category for Business Applications, a designation Microsoft awards to technical experts recognized for community leadership and real-world impact. His selection as an MVP reflects sustained, practitioner‑level influence on ERP architecture, implementation patterns, and outcomes in manufacturing and distribution. MVPs are trusted advisors to Microsoft product teams and the broader customer community, contributing insight drawn from live environments, complex tradeoffs, and execution at scale.

We spoke with Dag about what the recognition represents, the challenges he’s focused on addressing, and why many supply chain leaders are framing the AI conversation incorrectly.

Microsoft named you an MVP in AI ERP. What does that mean to you?

The MVP designation is a recognition of my contribution to the Dynamics 365 community at large, whether public speaking, blogging, or assistance with troubleshooting. I’ve spent 25 years using Dynamics 365 in complex manufacturing and supply chain. In many ways, the challenges I saw 25 years ago have not changed: Companies are generating more data than they can act on, and the gap between what they could know and what they do know is costing them. Many times that leads to the need to modernize their ERP and invest in data consolidation, reporting, and analysis.

You have talked about being in the business of replacing Excel. Why does that problem still exist?

People still rely on manual entry and Excel files because companies grow faster than their systems do. You start in Excel. It works. Then the data outgrows it and you’re running a business on spreadsheets that only a handful of people know how to use.

I still see demand planning happening in Excel because world-class systems run complex business logic requiring high data quality and potentially multiple data sources. Then convincing people to let go of their spreadsheet requires a change in process and trust in the new technology, both of which take significant time.

For the above reasons, the real challenge is not the technology. It’s convincing people to let go of the spreadsheet they’ve used for years. And if you don’t have all that data connected, you will not realize the full benefits of applying AI.

What problems are you most focused on right now?

I’ve been most focused on three things.

The first is helping organizations deploy AI in a way that solves a real business problem. There is a lot of enthusiasm around AI right now without enough discipline around where it creates value. Our Smart Sourcing Agent is a good example of what that looks like in practice. You give it your supplier data, your ESG requirements, your delivery constraints, and it makes a sourcing recommendation that a procurement team would have spent hours reaching manually. The goal is not AI because your competitors are using it or because agents are supposed to be cutting edge. It is AI that removes a specific burden from a specific person and speeds up the process.

The second is getting more out of Dynamics 365. Most organizations are not fully utilizing what they already have. Through our Managed Services practice, we work with clients to close that gap, whether that means optimizing configurations, improving adoption, or building out capabilities the business needs but has not yet activated.

The third is supply chain visibility. If you don’t have a connected picture of your supply chain, you cannot plan effectively, respond to disruption, or apply AI in any meaningful way. Well-architected technology, like Microsoft Fabric, brings that data together so teams have a single, trusted view of what’s happening across their operations without waiting on manual reporting… it’s too late by then anyway.

Why do you think there’s so much talk of supply chain AI failing?

Today supply chain leaders are starting with the tool instead of the problem. I see a lot of companies setting up AI tiger teams or centers of excellence that are essentially AI looking for use cases. That is backwards. Start with the executive-level metrics you want to move: inventory turns, return on assets, environmental impact… then ask what technology serves that goal. Sometimes it’s AI. Sometimes it’s better machinery. Sometimes it’s just cleaner data and some automation.

The other mistake is expecting AI to correct misalignment. AI multiplies whatever foundation you build on. If your data is fragmented, AI scales the fragmentation. If your scheduling system can’t talk to your machines, you do not have a smart factory. You have a loud, confused one. This is exactly what our Inspire Platform is designed to do: connect the data sources that feed every downstream decision, including AI.

And one overused term I’ve stopped using almost entirely: real-time. IoT data is real-time. An agent running every 15 minutes is not. The industry has used that term in so many contexts where it’s not actually true that it has become meaningless. Instead look for fast, agile, and accurate AI solutions. I’m very intentional about giving realistic use cases, not marketing.

What is the one thing you most want supply chain leaders to understand right now?

That the data you are not using today is costing you. Manufacturers and distributors generate more data than almost any other industry, and it is expected to increase fivefold over the next three years. Most companies are capturing less than half of their data and analyzing maybe ten percent.

If you knew what you could know, you would do what you should do. And think of how far behind you’re getting while competitors leverage data to drive real business value and competitive advantages.

But it’s not a technology problem. Successful organizations are consolidating data, standardizing processes, and focusing on business value. I encourage all supply chain leaders to focus on data quality and how business processes should be redesigned to take advantage of AI.