After setting up the grading parameters for Level AI, you can check your inputs and send the model to Allsite.ai to run. After the run is complete, you load the results back into Civil 3D. All three steps start from the AI Design panel of the Allsite.ai ribbon.
Running Service AI instead? See Running a Service AI Model. The Check Inputs, review-point, and run workflow described below applies to both tools, and you can run Level AI and Service AI together as one combined run: see Running Level AI and Service AI together.
Before running the model, validate it with Check Inputs. This extracts Civil 3D and Allsite.ai objects from your model and checks that they meet Level AI requirements, without submitting a run. Check & run performs the same checks and then sends the model to Allsite.ai, so running Check Inputs first is optional. If there are no fatal review items, meaning priority 9 review comments, the model is sent to Allsite.ai for processing.
Drawing Health Before Running:
Before running Check Inputs or a run, make sure your Civil 3D model is fully up to date.
Rebuild corridors (especially if any show “out of date” warnings).
Rebuild or update the surfaces used by your design and targets if any are out of date.
If the drawing becomes out of date after heavy edits, such as delete, undo, or move actions, carefully run AUDIT and -PURGE (RegApps) on a copy of the drawing to clean orphaned data and stale references.
Check Inputs
Click Run design… in the AI Design panel.
The Run design window opens. Tick Level AI — Grading. The three options below it, Run What-If scenarios if not viable, Include services and Existing connections only, are the Level AI settings of the same names (see Level AI Run Settings). A change you make to them here is saved to your project settings when you click Check Inputs or Check & run; Cancel discards it. Hover over or click a ? icon for a description, and click the cog to open all of the Level AI settings.
Click Check Inputs.
A progress popup appears, and a Civil 3D status pane opens. You can dock the status pane to one side.
Validation messages are displayed in the model. Check Inputs can generate review comments on the AS-REVIEW-INFO, AS-REVIEW-WARNING, and AS-REVIEW-ERROR layers. Use the Toggle Review Layers dropdown to turn Info, Warning, and Error messages on or off so you can quickly review model issues. More detailed information about validation errors and resolution steps is available here.
Click a validation point to see more information about the validation error.
All review points have a priority from 1, the lowest, to 9, the highest. Review points are color-coded by criticality, with red indicating the highest criticality and yellow indicating lower criticality. Higher-priority review items are more likely to affect the model run. Priority 9 review comments block the run and must be resolved before the model can be sent to Allsite.ai. Lower-priority comments may still affect the run and should be reviewed and corrected where necessary.
The Allsite AI Assistant provides an overview of review errors, lets you navigate to each error, and gives guidance on how to resolve it.
After Check Inputs or a run, the Assistant's card also carries a Settings Report section saying how many feature values differ from your project's settings, with a button that opens the full report. See Project settings, feature values and the Settings Report.
Review comments are also logged in the Civil 3D Panorama Event viewer.
Not all review points need to be addressed. Review items below priority 7 may affect the model, but they may also be acceptable depending on the design context. Review the items before and after a model run. Review items below priority 5 are informational only, and their layer is off by default.
Level AI Run Settings
The cog beside Level AI — Grading in the Run design window opens the Level AI settings. The cog in the AI Design ribbon panel opens the Level AI and Service AI settings together. The settings pane also opens when a run starts. Use it to configure settings for the run.
Setting | Description |
Run What-If Scenarios if Not Viable | If the design is not viable, Level AI will run scenarios to try to create a viable design, such as adding walls, changing lot grades, or increasing wall heights. If this setting is disabled, non-viable designs return a viability overlay, and no iteration fee is charged. |
Include Services | Includes sanitary and storm services in the Level AI analysis. If disabled, existing storm and sanitary connections are ignored in the Level AI run. If enabled, existing and proposed services are included using the pipe minimum grade and minimum cover settings from Allsite.ai. Level AI propagates the minimum grade up road corridors to help confirm that every site can be serviced. A “Services grade contingency” is added to allow for pipes following non-optimal paths up road corridors. |
Existing Connections Only | When Level AI runs, proposed pipes and structures in the model with invert levels set are ignored. Only existing services are included. |
Pre-Grade Analysis | Creates a pre-grade surface for cut/fill analysis. The pre-grade surface is defined as the proposed surface minus the pre-grade offset specified below, or the offset specified on parcel and road objects. |
Pre-Grade parcel offset Road corridor parcel offset | Default offset from the proposed ground surface for the pre-grade surface. This can be overridden on individual parcels and roads. |
Add Potential Walls | Specifies where potential walls are placed around lots. Potential walls are used in what-if scenarios and can be automatically added to a model to help make future runs viable. |
Draw Building FFL Feature Line | Specifies whether to output a feature line for the finished floor level at the top of the building pad into the proposed surface. |
Level AI Run
After reviewing and validating your model, open Run design…, tick Level AI — Grading and click Check & run. The inputs are checked again and, if there are no priority 9 review items, the model is sent to Allsite.ai for the AI model run. Smaller jobs may take about 30 minutes, while large projects over 100 acres may take more than 2 hours.
Click Run Status to check the run status, or click the Allsite.ai button in the ribbon to open the project page in Allsite.ai and view progress.
Loading results
When the run is complete, click Load results in the AI Design panel. It loads the results of the last run submitted from this drawing, whether that was Level AI, Service AI or both, and asks you to confirm first because existing design results are replaced. If the drawing's last runs were started from the earlier separate Level AI and Service AI buttons, it asks which results to load.
To read the results, see Interpreting Level AI Results.
Running Level AI and Service AI together
Tick both Level AI — Grading and Service AI — Networks in the Run design window to run them as one combined run. Level AI runs first, and Service AI then designs the networks on the grading Level AI has just produced, so there is no need to load the Level AI results before Service AI runs.
In a combined run, three Level AI options are always on, so they show ticked and greyed out:
Include services, so the grading allows for servicing. Level AI carries the pipe minimum grade and minimum cover up the road corridors, which keeps the levels it designs serviceable by the storm and sanitary networks Service AI then lays.
Existing connections only, so Level AI grades to the existing connections and ignores any proposed pipes already in the drawing, which Service AI is about to redesign.
Run What-If scenarios if not viable, so that if the drawn layout cannot grade, Level AI tries scenarios such as adding walls or changing lot grades instead of stopping, leaving Service AI a grading to design on.
See Level AI Run Settings for what each setting does. This applies to the combined run only: your saved choices for Level AI-only runs do not change.
Check Inputs and Check & run check the inputs for both tools at once, and the run needs enough credit for both. Service AI's own requirements, such as selected networks, part families and an outlet for each network, still apply: see Running a Service AI Model.
When the combined run is complete, one Load results loads both the grading and the networks.