The purpose of this article is to provide context around SARA, the agentic AI parcel management service built into Share A Refund.
Overview
SARA is a comprehensive set of learning models that achieve human decision-making and problem solving in real-time for most of the services built into Share A Refund.
Merchants and Customers use SARA across a wide range of problem sets that would otherwise be achieved through manual computational analyses.
Background
The complexity inherent to parcel lends well to the features built into Share A Refund. For most, the general workflow of data aggregation, advanced computational workflows and distillation of advanced reports for the coordination of supply chain efforts for financial decision making is manual, or at best partially automated. For most, decision making is done at the summary level, and is limited to data representative to a limited data set. This represents the ways of old. Slow and limited.
Share A Refund set out to build, from the ground up, an AI-driven parcel solution, that set out to minimize bias and guard against misuse. To do this, the Share A Refund team built a tool set that integrates local data stores, including parcel cost data, invoice and payment information and customer-specific information into a service that delivers accelerated productivity, deeper insights and better decision making. All while, building from the core tenants of security and accuracy.
After a multi-year, collaborative effort, the team at Share A Refund produced an AI agent, SARA, that achieved the same requirements as targets for the largest shippers in the world: complex task handling on parcel data and event response reporting in various formats with real-time throughput. Prioritizing security and fidelity, Share A Refund achieved success.
SARA is generally available to all customers across the entire Share A Refund Ecosystem.
How to use Ask SARA
To invoke Ask SARA within your account, proceed through the following workflow:
- Log in to the Share A Refund dashboard.
- Navigate to the Ask Sara button.
- Open a new chat.
- Ask a question.
- Interact with the outputs.
Output types supported by Ask SARA
Users of Ask SARA by Share A Refund enjoy the following types of outputs:
- Embedded charts, graphs and tables that can be copied and downloaded.
- Exportable branded reports in CSV, XLSX and PDF types.
Advanced use for power users
Each chat is organized into a conversation, that includes a unique url that gives users the ability to revisit previous topics at a later date to re-run or modify prior reporting periods.
For consultants, such as supply chain administrators and parcel pricing experts, conversations are able to be included within key deliverables, both from a self-help way of use and through supporting material as outputs for management reports.
Given the fine-grain export controls built into Ask SARA, the coordination of inputs for the purpose of creation of reports is enabled through faster, more accurate computational workloads ran in real-time over the transactions included within the prompt queried.
Examples
The following section includes just some of the queries that can be ran over Ask SARA, and also provides context on some of the business rules that are currently built into the service.
Example 1
What is the total shipment count of shipping for each month in 2026 where service type is Post?
Output:
Example 2
What is the total savings by type for each month in 2026?
Output:
Example 3
Show revenue by type for each month in 2025.
Output:
Tips on failed queries
The joins of data on fields represented within each business rule makes the task of prompt engineering a process of scientific achievement, where a user’s ability to ask the right question starts with known what questions can be asked and then constructing a prompt to achieve the target output.
In the case of a prompt failing in practice, check the field names associated with the query, and ensure a near or exact match name is including within the prompt to achieve maximum results.
Benchmarking
Terms, Context, Carriers, Mappings with assignable priority are managed for a working collection of business rule. Each business rule is tracked and validated against query performance and tuned to achieve high levels of usability and consistency. Prior to release, performance shows a success rate of 100%.
Data schema
All data present within the Share A Refund dashboard is assigned to the models built into Ask Sara, with guards and controls in place to ensure advanced security persists over every conversation.
About Share A Refund
The most award-winning audit and recovery solution for parcel, supports thousands of companies, including government, large enterprises and small business customers throughout the world. Our mission is to democratize low cost shipping for all. Learn more about us and our company.
