A Problem with Generic ERP

There are three specific places where we identify gaps within the generic ERP.

First is the batch. This is the basic unit of cost, yield and traceability in food manufacturing, and generic ERPs were not built around it. Batches have production orders made for one input and one output, but food is not made that way.  Unlike a generic ERP, a purpose-built ERP system specific to food creates a lifecycle record that the system will actually track. Therefore, cost and traceability are recorded correctly from the beginning. 

Then there is cost. To be most effective, you should have an end-to-end solution that can reconcile buying/selling by weight or picking/counting by piece or case from the time you receive the product to the time you invoice the vendor. Generic ERPs can't do that on their own, so spreadsheets are built to address the gap, leading to gross inaccuracies and confusion at month-end. To avoid these financial discrepancies, this sort of information should live inside an operating system, not in Excel.

Explore More Automation

And finally, compliance. By July 2028, according to the FDA Food Traceability Rule, manufacturers with products on the Food Traceability List must submit a report to the FDA within 24 hours if requested. It has become a common scenario for companies we have been working with to take several hours during a trial request to produce this information because their information is living in disparate places within the system they have built. The workarounds they have put into place fail them time and again because the data needs to live in a responsive, purpose-built ERP system, not a generic ERP program relying on spreadsheets created by hand.

It is inevitable that generic ERP systems are going to show their true colors and ultimately fail when put to the test.  We have seen this for 20 years, and with that, food manufacturers should be aware that AI is not much different from an operational ERP application and should be treated the same way — trained to be purpose-built.

Getting AI Right

AI is being sold to the food manufacturing industry the same way ERP systems were years ago. It is the latest shiny tech in the room, and like ERP sales meetings of the past, it is a pitch of capabilities assessments and review of interactive demonstrations. However, there is something missing — the question of real-world compatibility: will the AI actually work with the customer’s existing data systems in place? Unfortunately, this vital pre-integration step leads to massive mishaps in the future when companies realize the AI is generic and was not purpose-built for their industry.

There are two major types of AI, stochastic and deterministic, and they have very different roles in the future of food ERP. 

Stochastic AI (like ChatGPT and other generative models) may not directly control core ERP processes, but it can enhance employee productivity, assist in decision-making and streamline everyday tasks. 

For example, stochastic AI can assist in drafting an email to a supplier about a short shipment. That’s useful, but it doesn’t go as far as changing any financial impact in the system. On the other hand, deterministic AI applies fixed rules to recalculate batch cost based on the actual weight of the shipment received and update the traceability record.

In contrast, deterministic AI is set to become deeply embedded in ERP systems. In the food ERP world, companies are dealing with traceability, costing, inventory valuations and money. They must be precise because a wrong number is a wrong number, and that is simply unacceptable. Therefore, in this context, using AI that is probabilistic isn't going to cut it. 

Integrating deterministic AI within ERP systems enhances automation directly inside the ERP itself, relying on established rules and logic. As a result, inventory signals can trigger earlier, variances surface automatically, and entire workflows can be managed by agents without human initiation. This streamlined process not only improves efficiency but also ensures accuracy, which is non-negotiable in the food manufacturing industry.

Let’s use a finance department as an example. Agents can perform multiple tasks that speed up and streamline operations. They essentially become extended members of your team, monitoring inbox communications, processing new invoices, generating purchase invoices, conducting month-end closing tasks, and providing detailed reports on what remains to be done by a human. It greatly speeds up the process because the tasks are being done right in the system, not by a person too busy and bogged down in spreadsheets to address them.

But the question remains, and it is one we frequently hear: Can agents do the work, and more importantly, can they be trusted? If something goes wrong because an agent operated outside of the established rules, it could have a devastating effect on a business. So how do we avoid that?  

Ensure the agent is not left to its own judgment. It must have guardrails around a defined set of operations specific to the task it is seeking to accomplish. If the agent finds it needs assistance, it should be programmed to flag a human for interaction, rather than the other way around. While agents make things quicker, faster and easier, it is important to maintain a human touch in their interactions. Make sure tasks are done right, check the agent’s activity and step in if something looks off.

Most professionals in operations have been managing this kind of oversight throughout their careers. The only change is that they are using a different tool to enhance efficiency. Just like a generic ERP in food manufacturing will never function as efficiently as a purpose-built ERP, the same holds true for AI. Agents must be purpose-built to understand the industry they are working in and the components that make up the engine, including lot records, catch weight, FEFO inventory and quality control. 

We encourage food manufacturers to ask hard questions of any vendor: Is the tool purpose-built for the food industry? Can it meet the real-world needs of the system you run on? That goes for both ERP and AI vendors with real expertise in the food industry can answer questions. Those who can’t will struggle to explain whether the tools they have were ever really built for the job.