# Automation discovery (Preview)

> Automation discovery capability that mines messaging channels, email, wikis, and business systems for repetitive work and produces a prioritized automation backlog.

Before you build a specific automation, you need to know what to automate. With a coding agent, you point it at your organization's messaging channels, email, wikis, and systems of record; it investigates how people actually work and returns a prioritized backlog of automation opportunities, each backed by evidence rather than opinion.

## What it does

The agent mines Slack or Teams channels, email patterns, wikis, CRM/ERP/HRIS systems, and issue trackers for manual work, repeated questions, stalled approvals, and swivel-chair data entry. It looks specifically for:

- **Single points of failure** — roles that are the sole responder for a process, pseudonymized by default unless you authorize naming individuals.
- **Proven replicable models** — an automation already working in one team that could extend to others. These are treated as the highest-value finding.
- **Strategic gaps** — automation opportunities tied to the company's documented priorities (revenue growth, cost reduction, compliance, and similar), drawn from public or internal strategy sources.

It produces a four-tier prioritized report — replicable models first, then single points of failure, then evidenced opportunities, then strategic gaps — with every finding backed by a specific source, metric, and affected role or team. On request, it also sizes the build effort for each opportunity (complexity band, estimated hours, contingency), using effort-estimation references you supply.

## Privacy and authorization

Before mining anything, the agent confirms you're authorized to analyze the systems and data involved, avoids private channels, direct messages, and special-category HR data unless you explicitly approve it, and asks about jurisdiction constraints such as GDPR or works-council requirements. Individuals are referred to by role or pseudonym by default.

## The skill behind it

This work is powered by the `uipath-automation-discovery` skill, which mines your systems, analyzes the findings, and hands off each opportunity to the UiPath skill that builds it — RPA, a Maestro Flow, an agent, a human-in-the-loop step, or an integration. You can see it in the [Skills catalog](skills-catalog.md).

## A typical starting prompt

A first request might look like this:

"Suggest the highest-value processes we could automate."

The agent will ask what systems it can access and how broad a scan you want before it starts mining.

## More prompts to try

- Run a deep-dive automation audit across Finance and Support, and flag any single points of failure.
- We already automated invoice matching in Procurement — find out if the same pattern shows up manually elsewhere in the company.
- Size the build effort for the top 5 opportunities in the report.

## Go deeper

For where discovered opportunities get tracked and prioritized as a delivery pipeline, see the [Automation Hub documentation](https://docs.uipath.com/automation-hub).