Spanish Version
How many times have we thought that automation in our JDE system had a limit? You reach that point where someone has to look at something, manually review a document, or make a subjective decision. That’s the wall many Orchestrator projects bounce off. It’s the endless cycle of "almost-automation."
Well, last Friday, representing Quistor at the Orchestrator workshop in Barcelona, I witnessed that wall come down. And not with theories, but with practical examples you can start using today.
Mark Herwegue and Paolo Borrielo from Oracle put us to work with something that sounded like science fiction but worked right before our eyes: teaching JDE to "see" and understand.
The Practical Case: Is this the correct resistor? Let the AI decide.
Imagine you receive a batch of electronic components, resistors, for example. Quality is key. One error here can shut down an entire production line. The usual process involves a person manually checking them, comparing color codes... a slow, error-prone job.
What we did was different:
The Photo: An operator simply takes a picture of the incoming resistor.
Orchestrator Jumps In: The Orchestration sends that image to an Oracle Vision (AI) service in the OCI cloud.
AI Works Its Magic: A pre-trained Machine Learning model analyzes the resistor's colors, calculates its value (ohms), and sends the data back to Orchestrator.
JDE Takes Control: Orchestrator receives the value, compares it to the expected value from the purchase order, and decides:
✅ If it matches: It automatically creates the quality test as "Passed" and performs the inventory receipt. Process closed.
❌ If it doesn't match (or the image is blurry): It creates the test as "Failed," reports the error, and puts the order in an inspection status for human review.
Goodbye to interpretation errors. Goodbye to quality control bottlenecks. And all orchestrated from the ERP you already know.
But what if the operator forgets to upload the photo?
This is where things got even better. Because technology is one thing, and the reality of day-to-day work is another.
They showed us the power of Stateful Orchestrators and a new while loop functionality. We created an Orchestrator that:
Sends a notification to the responsible party's mobile device with two links: one to upload the photo to the order's Media Object and another to "resume" the process.
The orchestration pauses. Yes, it calmly waits for the human to do their part.
Once the photo is uploaded, the user clicks the second link, and the flow continues, calling the image recognition AI I mentioned earlier.
And if they forget to upload it? Here's the novelty! When the orchestration resumes, it enters a
whileloop that checks if the attachment exists. If it's missing, it sends the notification again. No more having to chase people down!
What does this all mean for us?
What I saw in Barcelona wasn't just a simple demo. It was confirmation that the barrier between the ERP and Artificial Intelligence has been broken.
We're done talking about theories: AI is no longer an abstract concept. It's just another service (like a database or a server) that we can consume from Orchestrator.
True End-to-End Automation: We can now include human tasks and visual validations in our automated flows.
Orchestrator is more than just an integrator: It solidifies its role as the true operations brain, capable of directing not only JDE but also advanced cognitive services.
It was an incredible session, not only for the technology but for connecting with other consultants and clients who face the same challenges. We saw how these new Oracle tools provide us with real solutions to age-old problems.
Now I ask you: What is that one manual process in your company that you thought was impossible to automate, but which something like this could solve? Quality control, invoice validation, meter reading?
I'd love to read your ideas!