Web AppAI

Edex Food

Predict the rush before it hits.

Web AppAI

Edex Food

Predict the rush before it hits.

Scroll

00The short version

Edex Food came to us with a problem every delivery network has: by the time the dispatcher reacts to a demand spike, half the orders are late. They wanted an AI that called the surge five minutes before it happened.

01Overview

The brief

Edex Food is an operations dashboard for food delivery fleets, including live metrics on rider performance, kitchen wait times, and order throughput. We layered an AI demand-prediction model on top so the dashboard can pre-position riders before surges hit. We designed and built the dashboard plus the AI integration over ten weeks.

02Challenge

What we solved

Live operations dashboards are notorious for becoming dumping grounds for every metric the team can think of. The result is a dashboard nobody actually uses during a shift because nothing is prioritized. Edex needed a dashboard built for the dispatcher's actual decision rhythm: every five minutes, what changed, what should I do.

03Approach

How we got there

We rebuilt the dashboard around the dispatcher's five-minute loop. The home view surfaces only what's actionable right now: predicted demand surges, riders to reposition, and kitchens running long. Historical reporting gets its own surface for after the shift. Claude powers the prediction layer using historical order data, weather, local events, and time-of-day signals.

04Solution

What we shipped

The web app gives ops managers a single live view of the entire delivery network with a five-minute prediction horizon overlaid. Riders that should be repositioned get flagged automatically. Kitchens with rising wait times surface at the top. End-of-shift reports are generated automatically with summaries of where predictions hit and missed.

05Results

18%Reduction in late deliveries

22%Less idle rider time via AI predictions

12Dark kitchens managed on the platform

18%Reduction in late deliveries

22%Less idle rider time via AI predictions

12Dark kitchens managed on the platform

05Outcome

What changed for the client

Edex cut late deliveries by 18 percent within the first month of using the new dashboard. The AI prediction layer reduced idle rider time by 22 percent, which translates to better unit economics on every order. Twelve dark kitchens are running on the platform.

FAQCommon questions

Frequently asked.

Questions that come up most often when companies reach out about projects like this.

Yes. Edex Food, LegalPal, AboutWork, and Chockablock all use Claude in different ways. For operations products like Edex we use Claude for prediction and natural-language summaries on top of structured data, paired with a model evaluation harness so the team can measure accuracy over time.

We anchor the design to the operator's actual decision rhythm. For Edex, that's a five-minute loop. For Big Gym, it's a daily loop. For Capisec, it's a per-incident loop. The dashboard surfaces only what's relevant to the current loop and pushes everything else to a separate reporting surface.

Most AI features are gated by data, not by model integration or UI. If the client has clean operational data, a Claude-powered feature can ship quickly under AXI Launch.

AXI Launch for the initial build. AXI Automate for ongoing AI model tuning, prompt iteration, and adding new agentic features over time. Most clients combine both: a Launch project to ship, then Automate to keep evolving.

Work with us