Intelligent Automation Program for an Italian Multi-Utility

From 2020 to 2025 we built and ran a managed automation programme of 20+ UiPath robots on the full UiPath Automation Cloud, owned by a Centre of Excellence inside the company, with monitoring, reporting and quarterly leadership reviews. Four of its flows show the range.

1. Security alert analysis

• Every alert read automatically. The robot collects the Darktrace alert reports from the mailbox they arrive in, unpacks the attachments and parses each alert: model, severity, source and destination IP, port and link.
• One queue item per alert. Alerts go onto an Orchestrator queue, so each is checked, retried and logged on its own.
• Threat intelligence on every IP. External addresses are checked against VirusTotal and IBM X-Force Exchange for recorded threats and country of origin, with a whitelist for known-safe addresses and a fallback geolocation service.
• Clear escalation rules. An alert is escalated if the source or destination is outside the EU, if either has recorded threats, or if it matches a watched Darktrace model such as denial of service, cryptomining, BitTorrent, Pastebin, Tor downloads or FTP.
• Reports for the administrators. Escalations reach IT security as formatted HTML emails; every alert is logged in daily processing and alert reports, with a separate exception log.

2. Telecom invoice collection

• Five operators, one robot. The robot logs into six portals of five telecom operators with credentials held in Orchestrator, for each company account and line type, and downloads invoices and technical documentation as PDF and CSV.
• Nothing downloaded twice. It checks what is already filed, handles zipped downloads and renames files to one convention.
• Filed where finance works. Documents went to the finance document repository and, later in the programme, into ServiceNow.
• Runs on the billing cycle. Monthly, or every other month per portal, with a run report and Italian-language notification emails to the process owner, including any business exceptions.

3. Fuel-card checks

• Transactions downloaded. The robot downloads every refuelling transaction on the corporate fuel cards from the supplier's customer portal.
• Four automatic checks. Fuel lines only, sorted by plate and date, then checked for wrong mileage, mileage delta between fill-ups, average consumption and price.
• Anomalies to HR. A full checked report and a separate anomalies report are saved and emailed to the owners.

4. Fuel cost allocation

• Statements read automatically. The robot reads the PDF fuel statements from two suppliers and extracts each vehicle plate and amount.
• Costs mapped to the right company. Each transaction is matched to the existing cost data and written onto the template of the group company that owns the vehicle, across seven companies.

Across the programme

• A Centre of Excellence inside the company. We set it up to run the programme: pipeline, standards, ownership and the link to leadership.
• The full UiPath Automation Cloud. Orchestrator queues, assets and credentials; Action Center for human-in-the-loop steps; Document Understanding with AI for document-based processes.
• Built to one standard. Robotic Enterprise Framework, configuration files, business versus system exceptions, exception logs and screenshots, test harnesses.
• Monitoring and governance. A monitoring system with reporting, and quarterly check-ins with leadership to review the programme and decide what to improve next.

Technology

For five years the programme ran as a production system, not a collection of one-off robots: 20+ automations, one standard, one owner inside the company.

• 20+ automations in production across IT security, finance and HR, from 2020 to 2025.
• A Centre of Excellence inside the company owning the programme.
• Every Darktrace alert checked against two threat-intelligence sources, with escalations and reports for the system administrators.
• Monthly invoice collection from five telecom operators taken off staff entirely.
• Every fuel-card transaction checked four ways, with anomalies reported, and fuel costs allocated across seven group companies.
• A programme worth several hundred thousand euros over its life, monitored and reviewed quarterly with leadership, until the client moved its automation to Microsoft Power Automate in 2025.
• Scope, not savings. Time, cost and ROI were not measured; the figures on this page describe the programme.

The risk most teams miss

Twenty robots are not twenty projects; they are a production system. Portals change their screens without notice (the telecom robot's history is a long list of screen and pop-up fixes), threat-intelligence APIs rate-limit, mailboxes fill up. Without monitoring and an owner, a robot fails silently and the business finds out when an invoice is missing or an alert went unread. Here the programme ran through a Centre of Excellence, on central monitoring, per-item queues and exception reporting, with quarterly reviews where leadership saw what was running and what needed fixing.