Shift Receipt

## Inspiration

Your paycheck should not require an investigation.

In Los Angeles, retail workers may be entitled to additional pay when their schedules are changed with little notice. The problem is that many workers may not know when a schedule change qualifies, how much it may be worth or what to do next.

Employer scheduling systems already help businesses track these changes. We wanted to build something from the worker's perspective.

That became **ShiftReceipt**.

## What it does

ShiftReceipt is an AI agent that turns schedule changes into something workers can understand and act on.

Workers upload their original schedule and their updated schedule. ShiftReceipt compares the two versions and identifies changes such as reduced hours, added hours, moved shifts, cancelled shifts and short rest periods between shifts.

The agent then generates a receipt showing what changed, the rule that applies and how much the worker may be owed.

But we did not want to stop at detection.

With the worker's approval, ShiftReceipt prepares Los Angeles' official Notice to Cure and sends it to their manager.

**Find the change. Calculate the impact. Help the worker act.**

## How we built it

ShiftReceipt runs as a Fetch.ai uAgent inside ASI:One.

Gemini reads uploaded schedule screenshots and converts them into structured shift data. From there, our Python rules engine compares confirmed schedule versions and performs the actual calculations based on Los Angeles' Fair Work Week rules.

We intentionally separated AI interpretation from the money calculation. Gemini reads the schedule while tested Python code calculates the result.

SQLite stores schedule versions, user information, answers and notice history. When the worker chooses to move forward, ShiftReceipt fills the official Notice to Cure PDF and sends it through Gmail.

This creates one conversational workflow from screenshot to receipt to action.

## Challenges we ran into

One of our biggest challenges was making sure ShiftReceipt was solving a problem in a way that did not already exist.

Scheduling platforms already help employers manage schedules and understand labor costs. There are also tools that broadly address wage issues.

We kept asking ourselves one question:

**Why would a worker actually use ShiftReceipt?**

That pushed us beyond creating another calculator or dashboard.

ShiftReceipt is designed from the employee's perspective. It preserves previous schedule versions, compares changes, explains why something may or may not qualify and helps the worker take the next step.

Another major challenge was creating something users could trust. Because we are dealing with people's pay and workplace rights, we did not want an AI model guessing how much someone may be owed.

That is why Gemini handles schedule extraction while deterministic Python code handles the calculations.

We also wanted the agent to act on behalf of the worker without taking control away from them. ShiftReceipt never contacts a manager automatically. The worker reviews the result first and explicitly chooses when to send the notice.

## Accomplishments that we're proud of

We are proud that ShiftReceipt is more than a chatbot that explains a workplace rule.

It completes a real workflow.

A worker can upload schedule screenshots, see exactly what changed, understand why it matters, receive a calculated receipt and then take action from the same conversation.

We are also proud that ShiftReceipt can determine when a worker may be owed **$0**.

The goal is not to manufacture violations. If an exception applies or a schedule change does not meet the requirements, ShiftReceipt explains why.

Most importantly, we built the product from the worker's perspective instead of the employer's.

## What we learned

One of the biggest things we learned is that using AI does not mean AI should make every decision.

Gemini is useful for turning an unstructured schedule screenshot into structured information. However, when money and workplace rules are involved, predictable and testable code is more appropriate for the calculation.

We also learned that detecting a problem is only part of a useful agent experience.

A strong agent should help move the user toward an outcome.

ShiftReceipt helps answer three questions:

- **What changed?**
- **Why does it matter?**
- **What can I do about it?**

We also learned how important user trust is when an agent is acting on someone's behalf. The worker remains involved throughout the process and decides when the agent should take action.

## What's next for ShiftReceipt

Next, we want ShiftReceipt to continuously track schedule versions across multiple weeks so workers can build a history of schedule changes over time.

We also want to integrate directly with scheduling and payroll platforms so workers do not always have to rely on screenshots.

On the employer side, ShiftReceipt could eventually help managers understand the potential impact of a schedule change before it is published.

Long term, we see ShiftReceipt expanding beyond Los Angeles to other cities with predictive scheduling protections.

Our vision is simple:

**Give workers an agent that helps them understand what changed, what it could mean for their pay and what they can do next.**

Built With

• agents
• agentverse
• api
• asi:one
• fetch.ai
• gemini
• generative
• gmail
• image
• ocr
• protocol
• python
• rest
• sqlite
• uagents