In short:
- Three collection methods coexist: agent, agentless, network discovery by scanning.
- The agent returns the most complete software inventory and works outside the corporate network.
- Network discovery is the only method that detects unmanageable equipment.
- A real estate almost always combines all three, at different collection frequencies by device type.
Why manual inventory does not hold
An IT estate moves constantly: workstations replaced, software installed on user initiative, virtual machines created then forgotten. A manually maintained IT inventory is wrong the day it is published. The question is therefore not whether to automate, but by which method.
What the inventory must return
At a minimum: machine identification, hardware characteristics, operating system and version, installed software and versions, associated user, date of last return. That last field is the most neglected and the most useful, since it alone distinguishes a device that has disappeared from a device that is simply switched off.
The three collection methods
| Method | Principle | Strengths | Limits |
|---|---|---|---|
| Agent | Program installed on the machine | Complete software inventory, works off the internal network | Agent deployment and maintenance |
| Agentless | Authenticated remote connection | Nothing to install, quick to set up | Machine must be on and reachable, less depth |
| Network discovery | Scanning of address ranges | Detects the unknown and unmanageable equipment | Shallow information, false positives |
The agent remains the reference on workstations and servers, because it is the only one reliably returning software inventory, which licence tracking depends on. Agentless collection is attractive for its lightness, but a laptop switched off at scan time is never inventoried.
What only network discovery provides
- Printers, switches and access points, on which no agent can be installed
- Undeclared equipment, added without going through the purchasing process
- Discrepancies between what finance records and what is actually plugged in
It is precisely this detection of the unknown that justifies scanning, even in an estate already well covered by agents.
Building the full chain
- Deploy the agent on workstations and servers, with a daily return
- Schedule a weekly network scan across all address ranges
- Reconcile both sources into a single repository, tagging the origin of every field
Source tagging gates everything downstream: without it, feeding a CMDB cleanly is impossible, as is knowing which value takes precedence. Inventory is only one brick of the setup, day-to-day use happening from the service desk, and the choice of collection tool being often dictated by the ITSM platform in place.
Frequently asked questions
How do you automate the inventory of an IT estate?
Three methods coexist: an agent installed on the device, which returns the most complete inventory and works outside the corporate network, agentless collection over an authenticated remote connection, lighter to deploy but limited to machines that are reachable and powered on, and network discovery by scanning, which detects unmanageable equipment such as printers or switches. Most estates combine all three, with agents covering workstations and servers and network discovery filling the gaps.
Does every device need an agent?
Not every device. An agent is justified on workstations and servers, where software inventory depth matters. Network equipment and peripherals belong to scan-based discovery.
How often should inventory collection run?
A daily return for active workstations and a weekly network scan is a common rhythm, enough to keep discrepancies detectable.