View case →CoolRunning AI: plain-English answers from delivery data.
An AI assistant for ice companies on Cool Running: ask about your own sales and routes, and get a table, a chart or a CSV back.
- Client
- Cool Running Software
- Country
- United States
- Stack
- React
- Tailwind CSS
- Node.js
- Express
- PostgreSQL
- pgvector
- SQL Server
- Google Gemini
- AWS
The results.
- 6AI tools: data, charts, weather, help and sources
- 0write access: every query is read-only
- 15 daysof weather forecast to plan ice demand
- 3 to 5insights a month for every licence
The challenge
Cool Running Software runs routes, deliveries and invoicing for ice companies in the United States. Each company's data sits in its own warehouse.
Getting an answer out of that data meant asking for a report or building a spreadsheet. Cool Running wanted owners and dispatchers to just ask: which stores are slipping, how this summer compares with the last one, and what next week's heat means for orders.
What we built
Ask in plain English
- Sign in with the Cool Running account, pick a licence and ask.
- The answer streams in with its reasoning, then a sortable table, a chart or a CSV download.
- Conversations kept by licence, with PDF export.
Safe by design
- Google Gemini writes the query; the server, not the model, chooses the database.
- One read-only SELECT at a time, with caps on rows, size and time.
- Every query logged for audit.
Knows the business
- Weather history and a 15-day forecast, because ice demand follows the heat.
- How-to answers drawn from Cool Running's own blog and training videos, with sources.
Comes to you
- Each month the assistant writes 3 to 5 insights for every licence.
- Sent by email to each company's admins and shown on the home screen.
Run as a product
- Plans, seats, coupons and a billing API for Cool Running.
- An admin portal for users, licences, conversations, prompt versions with a sandbox, usage and audit.
Stack
- React
- Tailwind CSS
- Node.js
- Express
- PostgreSQL
- pgvector
- SQL Server
- Google Gemini
- AWS
Screens.
(opens the full image in a new tab)
(opens the full image in a new tab)
(opens the full image in a new tab)
(opens the full image in a new tab)
(opens the full image in a new tab)
(opens the full image in a new tab)
(opens the full image in a new tab)
(opens the full image in a new tab)
(opens the full image in a new tab)
(opens the full image in a new tab)
In their words.
MindCore delivered our QuickBooks API weeks before it was due. They took the time to understand our unique business rules and handled adjustments with great patience. We couldn't be more pleased with their work and have already signed them for a much larger development project. Highly recommended for international collaboration.
Tony DicksonOwner, Cool Running SoftwareTake the case study with you.
A short PDF of this project to share with your team.
Want results like these?.
Tell us what you are building. On a 30-minute call we will say what we would build, how long it takes and what it costs.