26 Oct Is Manhattan Associates’ AI a game changer for the WMS world?

In February, at VLM’s winter event, I predicted four AI applications for the WMS. This month, at its European user event in Milan, Manhattan presented a series of AI innovations. Are these the game changers we have been waiting for?

The four applications are described in my March blog AI will be a game changer for WMS implementations. They are:

1. Answering questions
2. Controlling the warehouse
3. Configuring the WMS
4. Developing custom functionality

I called the last two game changers and the second a valuable addition. The first was already being introduced by several WMS vendors at the time.

Peter de Weerd covered the AI announcements in Warehouse Totaal (in Dutch). “These are not chatbots that only answer questions,” Chief Product Officer Brian Kinsella emphasized. That ticked off the first item on my list.

For the second item, AI that controls the warehouse, the report gives two concrete examples. A department falls behind schedule: AI agents look for available capacity and propose reassigning employees with the right skills. Or orders get stuck in the picking process: agents then suggest solutions, ranging from raising the priority of putaway tasks to making inventory available or moving trailers to a dock. Depending on the settings, a warehouse manager can approve the action or have certain actions fully automated. Manhattan stresses that the knowledge, intuition and judgment of experienced employees cannot simply be replaced by software.

Manhattan does add a warning to these examples: “If AI moves ten employees instead of three, or deploys employees without the right skills, the consequences can be disastrous.” That is why Manhattan works with predefined business rules and authorizations.

And this is where caution is needed. The term AI is used for very different techniques. A language model can interpret a problem and propose actions, but assigning staff under constraints requires a mathematical model. That is what operations research is for: models and algorithms that find the best solution within defined constraints. An AI agent can call such a model, but the report does not say whether that happens. Let the language model handle the conversation and let the math do the calculating.

The third application in my blog was AI that configures a WMS for you. Implementations are complex projects, and you depend heavily on the knowledge and experience of implementation consultants. It is a game changer if AI sets up the WMS while you share your requirements and wishes. Implementations become much faster, and changing the configuration later can become easier as well.

Manhattan announced Solution Design Studio, to be released in early 2027. In a test with an existing implementation document from a competitor, the system was able to configure about 90 percent of the requirements automatically, according to Manhattan. CEO Eric Clark says the standard implementation time of Active Warehouse can be reduced by up to 50 percent. If that holds up in practice, this is the game changer I predicted.

The fourth and final application in my blog was using AI to develop custom functionality yourself. In a standard WMS, you are limited to what the package can do. In practice, you almost always run into limits. Then you have to make concessions in your logistics or have custom functionality built. It is a game changer if you can add custom functionality yourself with AI.

Here Manhattan chooses a single integrated platform. You build custom functionality in Manhattan’s own ProActive environment. Manhattan sees Manhattan Active as the foundation for AI and states: “AI applied in a fragmented environment does not create intelligence, it increases complexity.”

I disagree. In my blog I argue for an open architecture in which the WMS forms the backbone and you can add your own software, a model we already know from ERP systems such as Microsoft Business Central and Odoo. An integrated platform can simplify internal consistency. But logistics optimization does not stop at system boundaries. The question is how freely you can connect other data sources, mathematical models and applications. Otherwise you remain confined to the standard.

Are these the game changers? For configuration, it looks that way; for custom functionality, Manhattan takes a different path. It remains to be seen how the AI applications work out in practice. I expect a lot from them. A pilot study at L’Oréal, presented at the event, showed a 14 percent increase in order picking productivity. The report does not explain which applications produced that gain or how the improvement was measured.

Be careful what you use AI for: for optimization problems, mathematics is the right tool. A language model is not a reliable optimization method in itself. Manhattan’s choice for a single integrated platform creates opportunities for WMS vendors that embrace an open landscape and free their customers from the straitjacket of standard software.