MEANS&METHODS LAB
ALL TOOL PAGES ARE DESK RESEARCH, NOT TESTED · ROUND 1 TESTING NOT STARTED
How we make money: nothing yet. Planned revenue never buys placement or rankings. Full disclosure

Home / Guides / Questions to ask before a facilities, CMMS or building-ops AI pilot

Questions to ask before a facilities, CMMS or building-ops AI pilot

A checklist for a first conversation with a facilities or building-operations AI vendor. These are our suggestions, drawn from our facilities category hub and our facilities, building operations and CMMS roundup. They are not findings: Means & Methods Lab has not tested any tool. Where a question mentions a vendor, it points to something that vendor publishes.

SUGGESTIONS, NOT FINDINGS. No tool has been tested by Means & Methods Lab. Vendor facts are desk research from vendor websites and are not verified. Updated 2026-10-15.

The questions (7)

  1. Q01Which decision are you buying for: work-order throughput, fault detection, HVAC energy, or portfolio workplace intelligence?

    WHY IT MATTERS · MaintainX and UpKeep center on CMMS work orders and PM. Willow and Facilio Connected Buildings emphasize cross-system faults and energy. BrainBox AI Control and Verdigris sit on HVAC optimization and electrical sensing. Metrikus sits on workplace and portfolio data.

    WHAT TO ASK · Write the one KPI the pilot will move (MTTR, kWh/sq ft, backlog age, or space utilization) before demos start.

  2. Q02How do construction as-builts, O&M manuals and asset tags get into the system?

    WHY IT MATTERS · Handover is where many pilots stall. Facilio and MaintainX describe asset and document workflows; Willow and Metrikus ingest live building systems. None of the pages we read claims a turnkey import of every GC closeout package.

    WHAT TO ASK · Bring a real asset register excerpt and one O&M PDF and ask them to load both in the session.

  3. Q03Does an AI recommendation cite the sensor, work history or document it came from?

    WHY IT MATTERS · Willow Copilot and Metrikus insights are described as grounded in building data. UpKeep Nova and MaintainX AI use work history and procedures. Ask how a technician verifies a suggested action before closing the work order.

    WHAT TO ASK · Open one AI suggestion live and ask them to show the underlying point, alarm or procedure.

  4. Q04What is autonomous vs advisory?

    WHY IT MATTERS · BrainBox AI Control and Willow Active Control describe autonomous response. ARIA and most CMMS AI features recommend or draft work for humans. Verdigris emphasizes validated limits for your own orchestration.

    WHAT TO ASK · Ask who can enable auto-writeback to the BMS, what the kill switch is, and how overrides are logged.

  5. Q05Which seats and AI features are in the tier you would buy?

    WHY IT MATTERS · UpKeep includes Nova credits by plan; MaintainX Assist is listed on Enterprise; Facilio Atom Suite is part of the AI pitch. Requester or view-only seats are often free on CMMS products.

    WHAT TO ASK · Price the pilot on the same seat mix you would run in production, including seasonal contractors.

  6. Q06What integrates with your BMS, ERP and existing CMMS — and what replaces them?

    WHY IT MATTERS · Metrikus says it sits above sensors and BMS. Verdigris lists BACnet/Modbus/MQTT/REST. MaintainX and UpKeep list IoT and API integrations. Confirm whether the vendor is a system of record or a layer.

    WHAT TO ASK · Name your BMS and ERP on the call and ask for a current integration sheet, not a roadmap slide.

  7. Q07Is building or tenant data used to train models outside your account?

    WHY IT MATTERS · BrainBox AI's ARIA FAQ describes closed-loop use of building data and SOC 2. Metrikus cites ISO 27001 and GDPR. Ask every vendor the same way in writing.

    WHAT TO ASK · Get retention, subprocessors and training-data policy before connecting a production BAS.

Bring to the demo

Sources

Tools mentioned (alphabetical, not ranked)

Side-by-side comparisons

Related

Means & Methods Lab has no commercial relationship with any vendor on this page. Rankings and placement are never for sale. See how we make money.