Zoomed Image
USA
800 691 9120
UK
01225 704844

AI Software Recognition

xAssets IT Asset Management User Guide
Software Licencing

AI Software Recognition

The AI Software Recognition Engine uses artificial intelligence to classify software titles discovered on your network. Unlike traditional pattern matching that relies on exact version strings and publisher names, the AI engine uses version-independent pattern matching to recognise software even when the installed title string differs from the expected format. This page explains how the AI recognition engine works, what it does automatically, and when manual intervention is still needed.

Prerequisites

How Traditional Recognition Works

Before understanding the AI engine, it helps to understand the baseline approach. Traditional software recognition works by matching the exact title string reported by the operating system against a database of known patterns. For example:

Discovered Title Pattern Result
"Microsoft Office Professional Plus 2019" "Microsoft Office Professional Plus 2019" Matched
"Microsoft Office Professional Plus" "Microsoft Office Professional Plus 2019" Not matched
"MS Office Pro Plus 2019" "Microsoft Office Professional Plus 2019" Not matched

This approach is brittle. Minor variations in how software reports its name — abbreviated publisher names, missing version numbers, localised titles, or rebranded editions — cause the title to go unrecognised.

How AI Recognition Works

The AI recognition engine adds an intelligent matching layer on top of the traditional pattern database. It uses the following techniques:

Version-Independent Matching

The AI engine strips version numbers, build identifiers, and edition suffixes from discovered titles before matching. This means a single recognition rule can match multiple versions of the same product:

Discovered Title AI Recognises As
"Adobe Acrobat Pro DC 2023.006.20380" Adobe Acrobat Pro
"Adobe Acrobat Pro DC 2024.001.20604" Adobe Acrobat Pro
"Adobe Acrobat Professional 11.0" Adobe Acrobat Pro

The actual version number is preserved in the discovery data — only the matching process ignores it. Reports and licence compliance calculations still use the full version information.

Publisher Normalisation

The engine normalises publisher names to handle common variations:

  • "Microsoft Corp." → "Microsoft"
  • "Microsoft Corporation" → "Microsoft"
  • "Adobe Systems Incorporated" → "Adobe"
  • "VMware, Inc." → "VMware"

Fuzzy Title Matching

When an exact match is not found, the AI engine applies fuzzy matching algorithms to find the closest known product. This handles:

  • Abbreviations (e.g., "MS" for "Microsoft")
  • Word reordering (e.g., "Professional Office" vs. "Office Professional")
  • Partial titles (e.g., "Acrobat" matching "Adobe Acrobat Pro")

Fuzzy matches are assigned a confidence score. Only matches above a configurable threshold are accepted automatically; lower-confidence matches are flagged for manual review.

What Happens Automatically

The AI recognition engine runs automatically during the discovery data loading process. When discovered software data is imported:

  1. Exact match — the system first attempts a traditional exact match against the recognition database.
  2. AI match — if no exact match is found, the AI engine attempts version-independent, normalised, and fuzzy matching.
  3. Automatic classification — high-confidence AI matches are classified automatically. The software title is assigned to the correct product in the licensing position.
  4. Flagged for review — low-confidence matches are placed in the Unrecognised Software list with a suggested match, allowing an administrator to confirm or reject the suggestion.

No manual intervention is required for high-confidence matches. The system learns from the recognition database updates provided by xAssets, so recognition accuracy improves over time.

What Requires Manual Review

The following scenarios still require manual review:

Scenario Action Required
Custom or internal software Create a manual recognition record (see Creating Software Titles Manually)
Low-confidence fuzzy matches Review the suggested match in the Unrecognised Software list and accept or reject it
Brand-new software not yet in the database Submit to xAssets for inclusion in a future recognition update, or create a manual record
Software with intentionally obscured names Create a manual recognition record

Viewing AI Recognition Results

To see the results of AI recognition:

  1. Navigate to the Software Dashboard.
  2. Review the Recognised Software and Unrecognised Software counts.
  3. Click into the Unrecognised Software query to see titles that could not be matched.

Titles that were recognised by the AI engine (rather than exact match) can be identified in the licensing position. The recognition source field indicates whether the match was exact or AI-assisted.

Configuration

The AI recognition engine is enabled by default and requires no configuration for standard operation. The following settings are available for advanced tuning:

Setting Description
AI Recognition Threshold The minimum confidence score for automatic classification. Matches below this threshold are flagged for manual review.
Publisher Normalisation Enabled by default. Can be disabled if your organisation uses non-standard publisher naming conventions.

Tips

Tip: Periodically review the Unrecognised Software list even when AI recognition is running. Some titles may be commercial software that requires licensing attention but was not matched by the AI engine.

Tip: When you create manual recognition records for internal software, classify them as "Free" or "Not Licensable" to prevent them from appearing in compliance reports.

Tip: Keep the recognition database up to date by applying software updates regularly. The AI engine's accuracy depends on having a comprehensive base of known products to match against.