To route pipes and ducts between two points in a Revit plantroom with AI, establish verified endpoints, capture the reserved maintenance zone and other keep‑outs, and include your approved sizes, slopes, and fitting rules; then request a small set of native proposal routes and independently review them against those requirements. AI accelerates option generation; engineers still choose and validate.
What an AI pipe and duct routing brief should include
The assistant works best when the team supplies facts it cannot infer: which connectors are valid, which areas are out of bounds, and which engineered rules apply. In practice, you provide:
• Verified endpoints: selectable elements with active MEP connectors or a named landing/connection zone for a not‑yet‑final target.
• Approved engineering inputs: system type, authorized sizes/ranges, slopes where applicable, and fitting preferences allowed by your standards.
• Constraint context: modeled corridors and explicit keep‑outs (including the reserved maintenance zone) that communicate intent.
In return, you ask for 2–3 native Revit proposal routes that attempt to honor the supplied constraints and rules. Treat the geometry as a coordination proposal. Unknown target elevation remains unresolved unless you explicitly supply a proposal range or landing zone that defines a vertical search band.
Prepare Revit plantroom context for a routing proposal
Keep the context in the model so proposals are auditable and repeatable across iterations.
• Corridors and no‑go volumes: model named 3D solids (or reference geometry) for “where routing is intended” and “where it is not,” and tag keep‑outs consistently.
• Maintenance access: represent the reserved maintenance zone as a continuous no‑go volume, not just annotations.
• Intent cues: add light‑touch preferences such as “favor perimeter” or “avoid above equipment face” as written instructions in your brief; do not rely on memory.
• Review copy: work in your team’s approved review model/copy method with dependencies verified so proposal elements do not double‑count in production quantities.
Prompts do not guarantee corridor or keep‑out enforcement. That’s why you embed checks and read the results; the assistant can flag exceptions where geometry deviates from your modeled intent.
Request pipe and duct route options for engineering review
This is a recommended team procedure, not product UI instructions.
1) Set the review context
• Create coordination views (plan/RCP as relevant and a 3D section box) dedicated to the plantroom.
• Reload and pin current links for the session; remember that pinning links does not freeze source files across teams.
2) Select endpoints and capture intent
• Select the source connector and either a target connector or a named landing zone. If the target elevation is not fixed, supply a proposal range; otherwise, leave it truly unresolved.
• State the permitted sizes/slopes/fitting rules from your approved standards.
3) Expose constraints in‑model
• Turn on corridor and keep‑out categories/filters; confirm names and tags are consistent for machine reading.
• Make sure the reserved maintenance zone is on and visible in the review views.
4) Ask for distinct native options
• Request two or three labeled strategies (for example, one that hugs a perimeter and one that rises earlier), and ask the assistant to stamp an instance‑level proposal flag for filtering in review views.
• Note that proposal parameters do not cause schedules to ignore elements automatically; use a review model/copy or explicit view/schedule filters to prevent duplicate quantities.
5) Generate and annotate
• Request native pipes/ducts and fittings using supported operations and verify their actual properties against your supplied rules, with an accompanying short summary (e.g., total length and fitting counts) produced for comparison.
• Have the assistant run the embedded checks and annotate exceptions in the option views.
6) Review and iterate
• Compare options in a short stand‑up. Pick a base option and request concise edits (shift around an obstruction, change a segment preference, or swap allowed fitting type) before final engineering validation.
For background on what current assistants can and cannot reliably automate in Revit, see What AI Can and Can’t Do in Revit Today: A Clear Guide.
Routing constraints to document before changing a Revit model
Embed these so findings arrive alongside geometry. Your team approves exceptions.
• Spatial and elevation
• Stay within named corridors and report any out‑of‑bounds segments.
• Do not intersect any modeled keep‑outs, including the reserved maintenance zone.
• If you supplied a proposal elevation range for a not‑yet‑final target, report the chosen value; otherwise, mark the target as unresolved.
• Fittings and constructability
• Use only permitted fitting types from your standards and flag any deviation.
• Summarize fitting counts per option so constructability can be discussed.
• System data and documentation
• Confirm System Type on all segments matches your request.
• Apply an instance‑level proposal flag for filtering; avoid editing shared type values to “label” proposals, since type edits affect all instances using that type unless your type strategy explicitly allows it.
• Coordination
• Check against active links for hard/soft clashes and list soft clashes separately with short notes for triage.
If you also route bundled cables, this mindset carries over; see Automating Data Center Cable Bundling and Routing with ArchiLabs for adjacent practices.
Plantroom routing example: compare two paths around an obstruction
Scenario (hypothetical): one pipe and one duct must traverse a tight plantroom from equipment on the south wall to services on the north wall. The duct discharge must reach a known riser opening; the pipe connects to a future coil whose exact height is not final.
• Constraints modeled: corridors defining intended paths, a reserved maintenance zone in front of the air‑handling unit, and a structural obstruction represented as a no‑go volume. A landing/connection zone is placed at the coil location; if the designer supplies a proposal elevation range, that becomes the search band; if not, the target remains unresolved.
• Options requested: “Perimeter Route” that follows the west perimeter as a preference and “Direct Route” that crosses earlier to shorten run length.
• What to review: placement inside corridors, any touches of the no‑go volumes, which fittings were selected under your allowed rules, and—only after modeling—the measured run lengths and fitting counts reported by the assistant. Do not infer success before the model is created and checked.
Check route geometry, connectors, and model context in Revit
A duplicated view is not frozen geometry or parameter state; proposal elements can still affect schedules unless you filter them or use an approved review‑model copy. Separate views do not isolate physical alternatives, and connected MEP design options come with limitations that your BIM lead should confirm for this task. Keep the alternative-management method and schedule scope explicit so reviewers know which elements belong to each proposal.
Using ArchiLabs Desktop for routing-related Revit tasks
ArchiLabs Desktop is positioned as an AI teammate alongside Revit and AutoCAD on Windows; it can create/update native model elements, route systems, and help produce documentation. Teams typically provide short screen recordings, annotated screenshots, and a selection set to convey intent. Integration details and specific tools vary by application version, plan, accounts, and permissions, so confirm the exact commands and reports available in your environment.
Prepare a clean review copy of a single plantroom, model corridors, keep‑outs, and a landing zone for any not‑yet‑final target, then draft a one‑page routing brief with the checks above and request two labeled proposals for your next coordination stand‑up.

Understand what Revit AI can do, where model access and source data limit results, and how to review geometry, parameters, and documentation.

Automate data center cable routing with ArchiLabs: clear logic vs. physical paths, safe bundling, rule-driven routes, BOMs, and impact analysis.