KT Sparks

Listed multi-utility · Italy · energy, water, waste and public services across five provinces

UiPath Automation Programme for a Listed Italian Multi-Utility

Nobody had time to read every security alert, and the one that mattered could be among them. A listed Italian multi-utility was reviewing network alerts, telecom invoices and fuel-card statements manually, and each review was a gap where a threat, an overcharge or card misuse could get through. Between 2020 and 2025 we built and operated a programme of 20+ UiPath robots. An in-house Centre of Excellence owned it, and leadership reviewed it every quarter.

UiPath Automation Programme for a Listed Italian Multi-Utility
Industry
Energy & Utilities
Function
Information Technology
Region
Italy

Results

20+
automations live in production
IT security, finance and HR on the full UiPath Automation Cloud. Figure from the delivery team.
5 yrs
in operation, from 2020 to 2025
Under monitoring and reviewed with leadership every quarter.
1
Centre of Excellence, in-house
Set up within the company to own the pipeline, standards and monitoring.

01

The challenge

The client is a listed multi-utility operating in five provinces of northern Italy. Its businesses span water, waste, electricity and gas, generation and district heating, public lighting and e-mobility. The group grew out of several local utilities and ran all of their back offices together, so work was split across many legal entities, suppliers and systems. A large share of it was manual, repetitive and driven by rules.

Where it hurt

  • Alerts left unread. Darktrace sent alert reports every hour. Looking up each source and destination IP in threat intelligence by hand takes a long time, and the dangerous alert is the one nobody reaches.
  • Five telecom portals a month. Staff had to download invoices and technical documentation from five telecom operators, for several company accounts, and file them internally.
  • Fuel cards leaking money. Someone checked corporate card refuelling transactions manually for wrong mileage, unlikely consumption and odd prices. Those are the places where errors and misuse go unnoticed.
  • Manual cost splits. Two suppliers sent fuel statements that someone had to divide by vehicle plate among seven companies in the group.
  • Nobody in charge of automation. With no Centre of Excellence, every robot would stay a one-off. Nobody would choose the next process to automate or confirm that existing robots still worked.

02

What we did

Between 2020 and 2025 we built and operated a managed programme running 20+ UiPath robots, built on the full UiPath Automation Cloud. An in-house Centre of Excellence owned it, with monitoring, reporting and leadership reviews every quarter. Four of the flows give a sense of its breadth.

1. Reviewing security alerts

  • No alert skipped. The robot picks up Darktrace alert reports from the mailbox where they land, extracts the attachments and splits every alert into its fields: severity, model, link, port, and the source and destination IP.
  • Each alert queued separately. Alerts become individual items in an Orchestrator queue, so every one is checked, retried and logged independently.
  • Every IP looked up. External addresses go to VirusTotal and IBM X-Force Exchange to find recorded threats and the country of origin. A whitelist covers addresses known to be safe, and a backup geolocation service steps in when needed.
  • Fixed rules for escalation. The robot escalates an alert when the source or destination sits outside the EU, when either one has recorded threats, or when the alert matches a Darktrace model under watch, for example FTP, Tor downloads, Pastebin, BitTorrent, cryptomining or denial of service.
  • Reporting to administrators. IT security receives escalations as formatted HTML emails. Daily processing and alert reports record every alert, and exceptions go to their own log.

2. Collecting telecom invoices

  • One robot for five operators. Using credentials stored in Orchestrator, the robot signs in to six portals belonging to five telecom operators, covering each line type and company account. It then fetches invoices and technical documentation as CSV and PDF files.
  • No duplicates. It checks what has already been filed, unpacks zipped downloads and renames every file to a single convention.
  • Filed where finance can use it. Documents were stored in the finance document repository and, in a later phase of the programme, in ServiceNow.
  • Timed to billing. It runs monthly or bimonthly depending on the portal. The process owner gets a run report and notification emails in Italian, business exceptions included.

3. Checking fuel cards

  • Pulling the transactions. The robot downloads all corporate fuel-card refuelling transactions from the customer portal run by the supplier.
  • Four automatic tests. It keeps fuel lines only, orders them by plate and date, and tests each for wrong mileage, the mileage gap between fill-ups, price and average consumption.
  • Anomalies sent to HR. Two reports come out, one listing every checked transaction and one with the anomalies only. Both get saved, and the owners receive them by email.

4. Allocating fuel costs

  • Statements parsed automatically. The robot reads PDF fuel statements from two suppliers and pulls out the plate and amount for each vehicle.
  • Each cost lands with its company. Every transaction is matched with existing cost data and entered into the template of whichever group company the vehicle belongs to, for seven companies in total.

Programme-wide

  • An in-house Centre of Excellence. We established it as the body running the programme, covering the pipeline, standards, ownership and the connection to leadership.
  • All of UiPath Automation Cloud. Queues, credentials and assets in Orchestrator, Action Center for steps that need a person, and Document Understanding with AI for processes built around documents.
  • A single build standard. Robotic Enterprise Framework, configuration files, a split between business and system exceptions, exception logs with screenshots, and test harnesses.
  • Governance and monitoring. Reporting from a monitoring system, plus quarterly sessions with leadership to review the programme and choose the next improvements.

Stack

LayerTools
PlatformUiPath Automation Cloud: Document Understanding with AI, Action Center, and Orchestrator queues, credentials and assets
DevelopmentUiPath Studio and Robotic Enterprise Framework, with activities for UI automation, web API, PDF, Excel and mail
Security and threat intelligenceDarktrace, IBM X-Force Exchange, VirusTotal API, a geolocation lookup service
Business systemsFinance document repository, ServiceNow, supplier and operator web portals
ReportingReports by HTML email, run reports, exception logs

03

The outcome

This was a production system for five years, not a set of isolated robots: 20+ automations, built to one standard, with a single owner in the company.

BeforeAfter
Alerts from DarktraceReviewed manually whenever time allowedEach alert checked in two threat-intelligence sources and escalated by rule
Telecom invoicesDownloaded every month from five operators' portalsGathered and filed automatically in line with billing
Fuel-card transactionsReviewed one line at a timeFour checks on every transaction, anomalies sent to HR
Fuel cost allocationDivided manually among group companiesAllocated automatically across seven companies
Automation ownershipNo oneA Centre of Excellence, monitored and reviewed by leadership each quarter
  • 20+ automations live in production in IT security, finance and HR between 2020 and 2025.
  • An in-house Centre of Excellence responsible for the programme.
  • Two threat-intelligence sources checked for every Darktrace alert, with escalations and reports going to the system administrators.
  • Staff no longer collect invoices from five telecom operators each month.
  • Four checks on every fuel-card transaction, anomalies reported, and fuel costs split across seven group companies.
  • Several hundred thousand euros in programme value across its lifetime, with monitoring and quarterly leadership reviews, until 2025, when the client switched its automation over to Microsoft Power Automate.
  • We report scope, not savings. Nobody measured time, cost or ROI. The figures here describe the programme itself.

What usually gets overlooked

Twenty robots add up to one production system, not twenty separate projects. Portals change screens with no warning, and the telecom robot's change log is mostly fixes for screens and pop-ups. Threat-intelligence APIs apply rate limits. Mailboxes run out of space. If no one owns and monitors the robots, one stops quietly and the business only notices when an invoice fails to arrive or an alert sits unread. In this programme a Centre of Excellence ran everything with central monitoring, exception reporting and queues that track each item, and leadership saw in quarterly reviews what was running and what needed attention.