Breaking Silos: How to Prepare Your Data for AI-Driven Operations

AI this, AI that, AI will solve everything. In some cases, that's happening. But many companies have trouble getting started or making progress because they don’t have a good grip on their data.

Many leaders underestimate the work it takes to get their data ready for AI, and the struggle is even more real in supply chains. Often, data is fragmented and stored in too many locations, data points are defined differently across systems, and companies trying to adopt AI have too narrow a focus, said Mike Mullen, COO of Bear Cognition, an AI software and solutions company for asset-based carriers and third-party suppliers.  

In a constantly changing business environment with nonstop advancements in AI, the prevailing wisdom is that the companies that learn to master the tools are best positioned to pull ahead. As more organizations wade into AI adoption and then tread water, Mullen made it clear there’s a way not just to swim but to thrive in this ocean of technology that can help leaders make decisions.

We asked Mullen how leaders can better prepare their data, break down silos and set AI initiatives up for success:

How can AI-driven operational intelligence close visibility gaps and reduce disruptions?

The first level is visibility of what’s going on. You have these siloed data sets out there: a warehouse management system tracking your inventory, a transportation management system managing all your trucks and where they're going, and your ERP, where your salespeople input all their orders into your billing. A lot of times, these are all separate.

If you can get all that data flowing into one spot, what’s the visibility? For example, if I’m putting out an ad to help sales reduce the cost of something to increase sales, how’s that going to impact my inventory inside the warehouse? Do I need to be prepared for it, which will then impact all the truckers I’ll need on the TMS side?

Getting all that data into one place and offering good visibility dashboards to the different teams is step one. Once you have that, you’re going to identify problems inside of that data and see where you need to focus. Then it’s the discussion of how am I going to solve that — is it a simple, human-driven process change? Or do I want to approach this from an AI standpoint and try to automate that workflow, or do something else on the tech side to help move it forward?

Why is there so much lift to get the data ready?

You're underestimating the complexity of the data itself. It's highly variable. There are different formats or inconsistent definitions that are happening amongst the different teams, perhaps. Without standardization, these integration efforts will stall, and it'll slow things down.

Second is the ownership piece. The data is sitting across multiple teams in logistics, finance and operations, and if there's no single group accountable for bringing it all together, it's going to lead to misalignment.

One other piece that can sometimes fail is that people will say the AI component is an IT initiative: “Let's just give it to the computer nerds inside the company and have them get this up and going.” In reality, it's an operational transformation right inside of your organization, and until there's clear ownership — you've standardized these definitions, and there's a direct tie to those business decisions — it's going to be tough to see it through.

What are the signs that it's working well?

What are the outcomes that you're looking for? Ultimately, that is where it should fall. You're saying, “I want to improve margins in this area,” or “I want to reduce inventory in this area.” That's your sign. Have you gotten to that point? Can you do it?

What are other stumbling blocks on the front end?