What to agree before an HVAC optimization pilot

1. Name the issue the pilot must address

Start with an observed problem, not “make the building smarter.” It might be HVAC continuing after occupants leave, repeated manual adjustments or a cost pattern that the team cannot explain.

Record when the issue happens, which spaces or systems are affected and what evidence is available. An operator report is useful context, but it should be distinguished from a trend showing that the equipment actually ran.

Before adding a platform, ask the existing controls provider what a repair, schedule change or commissioning exercise could resolve. Record what has already been tried and why it did or did not remain effective.

Your first decision: Is there a specific, controllable operating issue worth evaluating, or is investigation or repair the better next step?

2. Confirm who receives value and who has authority

The organization paying for the pilot may not pay the utility bill. The landlord, occupier, property manager and service provider may also have different authority over the controls.

Record these responsibilities explicitly:

• Economic owner: approves spending and confirms who retains any financial benefit.
• Facilities lead: defines operational requirements and coordinates access.
• Controls provider: confirms interfaces, permitted points and existing control behavior.
• IT or security lead: approves the connection, access boundaries and data handling.
• Measurement lead: agrees the comparison method, checks data quality and reports uncertainty.
• Pilot provider: documents the proposed intervention and its support responsibilities.

One person may hold several roles. What matters is that no responsibility is left implied. Do not assume tenant bill savings automatically become landlord income.

3. Build a useful evidence pack

Prepare what exists rather than delaying the discussion until every document is perfect. Mark missing or uncertain items so the team can decide whether they matter to the proposed scope.

Building and operation: occupancy schedules, operating hours, relevant plans, known comfort complaints and recent changes in use.

Equipment and controls: equipment list, current sequences, BMS details, available points, command permissions, sensor information and existing overrides.

Historical data: relevant trends, meter readings, timestamps, units, gaps and any changes to data collection.

Commercial context: utility bills, applicable tariffs, bill ownership, existing service responsibilities and restrictions on access.

Known interventions: repairs, recommissioning, tenant changes or other work that could affect the comparison.

The purpose is not to collect every possible data point. It is to establish whether the available evidence can support the proposed decision.

4. Separate connection from permission to control

A system may expose useful readings without allowing the commands needed for a pilot. Even when a command can be sent, the team needs to verify whether it takes effect and what other control logic can override it.

Ask the provider and controls contractor to document:

• Which points will be read and which may be changed.
• The equipment and spaces affected by each permitted change.
• Any gateway, sensor, controller or network work required.
• Who owns credentials and who can approve access changes.
• What operators can review and how actions will be recorded.
• Who supports the connection and local equipment if something goes wrong.

BACnet or Modbus availability alone does not settle these questions. Treat compatibility as a site assessment, not a logo on a proposal.

5. Agree operating limits and stop conditions

Your qualified engineer and controls provider should define the applicable comfort, ventilation, equipment and safety requirements for the building. Do not use a generic temperature range or a single air quality reading as a substitute for that review.

Before equipment control begins, agree how the actual installation will be checked for manual override, unreliable readings, connection loss and return to existing control. A connection failure does not necessarily remove a command already held by local equipment.

Record who can pause the evaluation, who is notified and who verifies the return to the approved operating state. Conditions for stopping should be understandable to the team that will receive the call, not only to the person who configured the platform.

This is a planning requirement, not permission to experiment with live safety systems. Any test affecting equipment needs the appropriate site approvals and qualified supervision.

6. Define what would count as a result

Choose the measurement boundary before evaluating savings. A whole building electricity bill and a meter serving one HVAC system answer different questions. Record which energy sources and equipment are included, and which are not.

The DOE FEMP guidance on measurement and verification describes several approaches. System level measurements can address a defined system. Whole building comparisons include other loads and need suitable treatment of conditions such as weather, occupancy and operations. Calibrated simulation can support a measurement method when it is tied to the relevant baseline and installed conditions. A training simulation by itself does not establish savings in your building.

Agree how the evaluation will address:

• Weather, occupancy and operating hours.
• Changes in tariffs or demand charges.
• Repairs and other work during the measurement period.
• Missing data, sensor problems and unusual operating events.
• Uncertainty and conditions the available data cannot represent.
• Comfort, ventilation and equipment requirements alongside energy and cost.

Do not select a universal pilot duration before examining the building and the comparison method. The period must be suitable for the conditions and decision being evaluated.

7. Review the commercial boundary

The cost and quote for an Energia AI pilot depend on the agreed site scope: integration work, additional devices if needed, commissioning, service, measurement and ongoing support. Ask for a complete scope covering these items before committing. Clarify who pays for extra work if an interface or sensor is unsuitable.

Agree who can export data, what remains available after the evaluation and what happens to control access at its end. Record any continuing obligations before comparing a pilot with a longer contract.

Evaluate financial value after the relevant project and service costs. Keep energy use, bill changes and peak demand distinct. A change in operating time can affect the bill without producing the same change in total consumption.

8. Use a written decision record

Copy these prompts into the pilot approval document. Each answer should identify the evidence or the person responsible for resolving it.

Operating problem: What happens, where, when and with what supporting evidence?

Alternative first step: What can maintenance, commissioning or the existing BMS resolve without this pilot?

Approved scope: Which equipment, zones, data sources and commands are included? What is excluded?

Responsible people: Who owns economics, site operation, controls, IT, measurement and provider support?

Access and integration: What is available now, what needs work and who accepts that work?

Operating acceptance: Which requirements must remain satisfied, and how will the team check them?

Pause and return: Who can stop the pilot, what triggers a pause and how is the approved control state verified?

Comparison plan: What is the baseline, measurement boundary, method and treatment of other changes?

Commercial scope: Which costs and responsibilities are included, and what obligations continue after the evaluation?

Final decision: What evidence would support extending, revising or ending the project?

What should the final review contain?

A useful final report connects the intervention to the evidence. It should identify what actually changed, which data were usable, whether operating requirements were met and what uncertainty remains.

It should also separate the control result from other changes at the site. Report a finding of no clear improvement as plainly as a positive result. A functioning connection, successful demonstration or attractive dashboard is not by itself a reason to expand.

There are three reasonable outcomes:

1. Proceed: the evidence supports the agreed operational and commercial case.
2. Revise: a specific unresolved issue justifies a defined additional assessment, not an indefinite extension.
3. Stop: the evidence, site fit or economics do not justify continuing.

Prepare one building for a useful conversation

You do not need a complete portfolio plan to start. Bring one operating issue, the available trends and meter data, and your engineer’s view of the existing controls.

Discuss an Energia AI pilot to review fit and agree what an evaluation would need to prove. For the underlying concepts, see our commercial HVAC optimization guide.

Sources and scope

This planning guide is published by Energia AI, an HVAC control provider. The process recommendations are a buyer checklist, not a claim of product certification or completed field deployment. Measurement concepts are supported by the linked DOE FEMP guidance. The importance of integration, control, support and economics is also reflected in the ASHB 2025 Smart Building survey and JLL 2025 Global State of Facilities Management report. Those surveys describe selected respondent groups, not every building or prospective Energia customer.