Building an AI tool around one person’s workflow

A technician needed a practical way to manage his parts inventory from his phone. Recording deliveries, updating quantities, and checking what needed replenishing had to fit around his working day. I built an application that lets him handle those tasks through voice, text, and images, with an inventory screen where he can inspect and manage the records directly. The starting point was a familiar problem: keeping records takes effort at the moment something changes, while the benefit often comes later. Knowing what is available depends on recording what arrives and what gets used. Making those updates convenient was central to the application’s purpose. ​

Recording a delivery​

A delivery can contain parts already held in stock alongside items being recorded for the first time. Adding it to an inventory means finding the matching records, increasing their quantities, and creating entries for anything new. The following walkthrough uses sample data in the application’s staging environment. I started with six existing items and a delivery note containing twenty.​

I attached an image of the delivery note and asked the application to add the arrivals. The request specified that pipe quantities were in metres and other quantities were in pieces. It also asked the application to preserve existing stock thresholds and apply category-specific thresholds to new items.​

Reading the list is one part of the task. Applying it requires working with the inventory already there. Ten arriving hoses need to be added to the two already held. A hose with a different length or connection needs a separate entry. The application reported six existing items updated and fourteen new items created. Its response listed the resulting stock quantities for review.​

The inventory screen showed the corresponding changes. The existing 50 cm flex hose increased from two to twelve units. The brass 90° elbow with 3/4-inch female connections increased from eight to twenty-eight. New hose variants appeared as their own entries.​

Asking what needs attention​

With the delivery recorded, the next question was simple: “Any parts running low?” I asked it by voice. The application returned the item at its low-stock threshold. The delivery note provided the incoming quantities. The inventory retained the updated records. A spoken question brought back the information needed to decide what to replenish. This gives the user a choice of how to work. An image can supply a list that would otherwise need entering item by item. Text can express instructions for handling it. Voice can make a quick question convenient from a phone. The inventory screen provides a place to browse and inspect the records whenever that is more useful.​

Building around a recurring task​

The useful connection here is between the way someone describes their work and the structured records that support it. A delivery arrives as a document. A question comes naturally as a sentence. The application connects those inputs to inventory actions and information. That pattern extends beyond managing parts. A small team, an independent professional, or someone running a personal project may have recurring tasks that involve gathering information, updating records, and checking what needs attention. A focused application can bring those steps together around how the person actually works. The starting point is to identify a task that happens often enough to matter and understand where the effort goes. In this project, that meant giving a technician a practical way to maintain and consult his inventory from his phone. Which recurring task would become easier if you could describe what you needed and have a tool carry out the relevant action? The answer doesn’t have to involve a company-wide transformation. ​

You can explore this approach yourself in the interactive demo on the Reaigent website.​

For more projects and practical applications of AI, follow Reaigent on LinkedIn.​