1. Open Autodesk Navisworks Manage and open or append the models you want to coordinate. Save the federation as an NWF or NWD so ClashWise can keep its local mappings and presets with a stable document identity.
2. Open Clash Detective, create or select at least one clash test and run it. ClashWise does not detect geometry clashes itself; it works with the results Navisworks produces.
3. Open the ClashWiseAI® ribbon tab and use the Login/Logout control to sign in. A new account starts a 14-day Pro trial without a credit card. Internet access is required for the first sign-in and periodic subscription verification.
4. Open Property Mapping. Assign each appended model to the correct source value or model property, choose its discipline and add level overrides where needed. Review the live sample and save the confirmed mapping. ClashWise will not silently infer an explicit source from a filename, category or AI result.
5. Open AI Models and download the recommended on-device model. The first download can be large. Keep the recommended model unless you have reviewed the hardware guidance and tested another model against a real clash.
6. Click Title Update. The selection-first run sheet shows available tests, selected totals, the naming format and example, output destination, scope, language and mapping readiness. Select the tests or results you want to process.
7. Choose where the generated text should be written: title, description or comments. Choose whether to process only new clashes or all statuses, and whether to preserve the original title/test name or include previously updated results.
8. Choose one of the six built-in naming standards, configure Mixed, or open the format manager to build a token template. Token templates are applied exactly as entered and are not translated by the AI.
9. Start the run. The native review list fills as titles are generated. Select a row to inspect its clash in the model; edit the proposal, regenerate it, reject it or retry a failure. Stopping generation keeps proposals already produced.
10. Click Apply only after review. Accepted titles are written in one transaction, so one Ctrl+Z reverses that apply operation. You can also restore the just-applied run, or later use Restore original titles… on the same computer.
11. To prepare repeatable coordination locally, use Run and Publish and its Search Set, Clash Test and grouping workbenches. Review source assignments and proposed native artifacts before creating them. Search conditions support explicit AND/OR logic and a model/location scope. The automated Coordination Setup workflow requires an eligible active or trial Pro, Business or Enterprise organization.
12. Cloud publication, Context Models and Wise require an eligible Clash Management plan and an internet connection. If you only intend to name and prepare clashes locally, leave cloud fallback disabled and stop before the publication stage.

Installation

• Download the ClashWise Starter installer from Autodesk Marketplace.
• Close every running instance of Autodesk Navisworks.
• Run the downloaded .exe and follow the prompts. Installation is per-user and does not require administrator rights. Files are installed under %APPDATA%\Autodesk\ApplicationPlugins\ClashWise.bundle.
• Start Autodesk Navisworks Manage. The matching 2024, 2025, 2026 or 2027 plug-in loads automatically and the ClashWiseAI® ribbon tab appears.

Uninstallation

• Close Autodesk Navisworks.
• Open Windows Settings → Apps → Installed apps (or Control Panel → Programs and Features).
• Select ClashWiseAI Navisworks Clash Naming Combined and choose Uninstall.

Local privacy boundary: with cloud fallback off, title-generation content remains on the computer. Sign-in, subscription checks, AI-model download, updates, support reporting and optional usage counts can still use the internet.

Hardware: 64-bit Windows, Autodesk Navisworks Manage 2024–2027, a recent multi-core CPU, 16 GB RAM and enough disk space for the selected AI model and cache are recommended. The current local engine runs on CPU.

Known limitations: restore history is local to the computer that performed the naming. Quality and speed vary with the selected model, model data and hardware. Wise-generated narrative is not part of the reduced installer.