Aqueduct questions, answered.

Milvian AI makes Aqueduct. Milvian AI is a product company: it builds and runs the platform, the data model, the AI and the digital twins, while partners survey sites, supply and install hardware and deliver facility services. Read more about the company.

Aqueduct models the physical system and proves results, where those tools collect, run or display. A BMS runs equipment sequences, a CMMS tracks work, an IoT platform moves readings and a dashboard charts them. Aqueduct balances supply against use in digital twins, compares use with a weather-normalized baseline, then routes or takes action. See what a smart building platform is.

Aqueduct can replace a BMS, but it does not have to. There are three ways in: keep the BMS and read its points over BACnet or Modbus, replace aged controls function by function, or specify Aqueduct at design stage for a new building. Which functions move is confirmed site by site during the survey.

No. Aqueduct works alongside both. Your CMMS keeps managing technicians, parts and work orders, while Aqueduct sends findings into it and reads closure back. Utilities keep operating pumps and valves from SCADA, and Aqueduct takes a read-only feed from SCADA, AMI and historians. See SCADA, AMI and historians explained.

Aqueduct works in any facility or network with meters, sensors or controllable equipment: hospitals, airports, office and retail portfolios, universities, data centers, manufacturing plants, dairy operations, water utilities and public buildings. Each site gets twins built block by block to match the systems it actually has, so a site without an RO plant shows none. Browse industries we serve.

Aqueduct finds leaks with a water balance: main supply should equal metered floors, fire system, RO feed and other known uses, and whatever remains is unaccounted loss. aQ WaterGuard recalculates that balance every 60 seconds in the site's digital twin and alerts when unaccounted flow rises, such as a steady overnight draw in an empty building. See leak detection.

Aqueduct can start from one main meter, but sub-meters narrow the search. The main meter shows that use changed overnight; meters on floors, lines, cooling towers or process loops show where. A site survey maps the meters already installed and the few lines that would close the balance. Read the water loss playbook.

Customers using Aqueduct have seen water savings of up to 50 percent, depending on how much was being lost before. The savings come from fixing leaks, overflows, stuck valves and wasteful schedules, and each one is reported against the site's own weather-normalized baseline rather than last year's bill. See aQ WaterGuard.

Yes. aQ WaterGuard tracks the sewage treatment plant's inlet, treated flow, efficiency and BOD out, and the RO unit's feed, permeate, reject, recovery and TDS. Reuse is then reported from meters instead of estimates, which matters for discharge permits and water disclosures. See water reuse, STP and RO.

Yes. For utilities, Aqueduct keeps a running balance per district metered area from inflow meters, customer meters and night flow, so loss points to a district and a date instead of a yearly figure. It reads SCADA and AMI without writing back, and operators keep running the network. See non-revenue water.

aQ EnergySmart measures electricity and gas from the main incomer down to feeders and equipment: kW, kWh, power factor, peak demand, voltage, current and carbon. Its electrical digital twin follows the single line, balancing grid, solar and generator supply against HVAC, lighting, plug loads, services and losses. Gas is reported in therms or cubic meters. See aQ EnergySmart.

Customers typically see energy savings of 10 to 35 percent, depending on the starting point and whether control is switched on. The largest gains usually come from HVAC schedules, ventilation matched to occupancy, peak demand and power factor, each measured against a weather-normalized baseline. See HVAC energy optimization.

Yes. Aqueduct shows when demand peaks and which loads drive it, then sequences loads so they do not all start together, for example refrigeration and grow lights. Batteries, solar or backup generation can also be scheduled against the peak. Changes run as recommendations or, with control switched on, as commands. See peak demand management.

Yes. Aqueduct compares what the supplier bills with your own meter and sub-meter data, flags gaps in consumption or rate, and separates a cost change caused by usage from one caused by a tariff change. Disputes then rest on interval data instead of a single monthly total. See utility bill validation.

Yes. Billing Management in aQ Core turns sub-meter data into tenant bills, shows the current cycle as it accrues and flags usage outside each tenant's normal range. On a shared meter, the landlord sees the total and every share, while each tenant sees only its own percentage. See tenant sub-metering and billing.

aQ AirSafe monitors CO2, VOC index, carbon monoxide, temperature and humidity, with pressure, light, motion and sound where needed. Each sensor speaks for a zone on the floor plan, and readings are compared with occupancy, so a CO2 rise in a full room is treated differently from one in an empty room. See indoor air quality monitoring.

Yes. aQ AirSafe matches ventilation to occupancy: more outdoor air for full rooms, less for empty ones. When CO2 breaches a target, it recommends a costed action, such as raising outdoor air while the weather outside is mild, and can apply it through Smart Controls or the BMS once control is on. See CO2 monitoring.

aQ AirSafe rates zones against the standard set for each site, such as ASHRAE 62.1, ASHRAE 241, WELL or England's Approved Document F. Each finding names the threshold behind it and explains why the rating was given. Exposure is tracked in zone-hours, because duration matters as much as peak level. See ASHRAE 62.1.

Partly. Where a pattern repeats, such as a cafeteria at lunch, nightly cleaning or a crowded meeting room, Aqueduct can anticipate it and adjust ventilation ahead of time. One-off events, such as a truck idling by a loading door, can only be caught as they happen, which is why monitoring runs continuously. Read the IAQ monitoring guide.

Fault detection and diagnostics (FDD) finds a fault that already exists and names its likely cause; predictive maintenance spots early wear so work happens before a fault. aQ PredictIQ does both: FDD on BMS points and meters, and prediction from vibration, sound and energy drift on critical machines. Read FDD vs predictive maintenance.

aQ PredictIQ covers air handlers, rooftop units, chillers, cooling towers, pumps, compressors and boilers. Each finding arrives with the affected component, the readings that triggered it and a likely cause, so a technician knows what to check before reaching the plant room. See aQ PredictIQ.

aQ PredictIQ sends each finding into the CMMS you already run as a work order with the cause and evidence attached. After the ticket closes, it keeps watching the same signals, and if the fault returns, the ticket reopens. Connections are scoped per project to the CMMS in use. See fault detection and diagnostics.

Usually only a few. Fault detection runs on the BMS points and meters a site already has. Vibration and acoustic sensors are added on critical rotating equipment, such as pumps and compressors that cannot be allowed to stop, where wear often shows before any alarm. See predictive maintenance.

Aqueduct controls equipment in two ways: LoRaWAN relay switches for simple loads such as exhaust fans and lights, and Modbus or BACnet through a field device network controller for air handlers, rooftop units, chillers and VAV boxes. Field devices also send telemetry back, so every controlled unit adds data. See controllers and actuators.

No. Aqueduct acts only where control is enabled, with owner approval, and within the limits the owner sets. Every command is written to the device's command history, which shows whether it executed, and lockout warnings block new commands while one is still running, protecting equipment from short cycling. See trust and security.

No. Every optimization can run as a recommendation that your team or BMS applies, and the saving is measured the same way either way. Direct control applies schedules and setpoints faster, so a site can begin with recommendations and switch control on later, one device or zone at a time. See AI optimization.

Occupancy comes from sensors or from a schedule, per building, floor or zone, and it can trigger automation rules. A rule might drop ventilation to its minimum when a zone empties, or raise it when CO2 climbs in an occupied room. Rules use if and else logic on any metric, with a severity for each alert. See HVAC energy optimization.

A control cluster is a group of devices and sensors that Aqueduct operates and analyzes as one system. A rooftop unit plus a zone sensor is a typical cluster: the sensor's readings steer how the unit runs, and a pressure reading from a linked sensor can raise a filter change alert. See replacing BMS functions.

Aqueduct's agents work in the background: they watch live telemetry and the digital twins, work out what is wrong, and either propose a fix or apply it through Smart Controls. Each action stays inside rules the owner sets, is logged, and is checked against the data afterwards. See agentic AI in Aqueduct.

aQ Assistant is a multi-agent assistant that answers questions about any site, zone or device in plain language. It knows which page you are viewing, queries the platform's data through tools, and returns analysis such as consumption, peak demand and week-on-week trends. See aQ Assistant.

Aqueduct uses AI as the method for outcomes, not as a product on its own. Regression models set weather-normalized baselines, diagnostics rank faults by likely cause, optimization rules tune ventilation and schedules, and aQ Assistant answers plain-language questions. Each result traces back to the readings behind it. See AI optimization.

aQ Assistant answers only from the platform's own data, so an answer is as good as the readings behind it. When coverage is thin, for example days when a meter reported for part of the day, it warns that comparisons are unreliable instead of presenting a confident number. Device health in aQ Core shows which sensor caused the gap.

Usually not at first. Aqueduct starts with the meters, sensors and BMS points a site already has, provided they send telemetry, and adds sensing only where the site is blind, such as an unmetered cooling tower or floor. A site survey lists the gaps. See the hardware Aqueduct works with.

Aqueduct reads LoRaWAN, BACnet, Modbus, MQTT, M-Bus and pulse outputs. Wireless sensors and relays usually join over LoRaWAN; air handlers, chillers and meters over BACnet or Modbus through a field device network controller; IP devices over MQTT. CMMS, SCADA, AMI and historians connect as systems of record. See integrations.

No. Milvian AI does not make sensors, and hardware is not its core business. Aqueduct is device agnostic: meters, sensors, gateways and controllers are chosen per site from established makers, and partners supply and install them. Anything that sends telemetry can be onboarded. See hardware partners.

aQ Core is the data model under every Aqueduct product: sites, floors, zones and devices, with ingest, storage and device health. Water, energy, air, maintenance and aQ Assistant all read from the same store, so a reading is collected once and reused. Each device shows its connection rate and battery or capacitor voltage. See aQ Core.

Yes. Pulse outputs connect through a pulse counter, and a meter with no output at all can be read by a camera reader that photographs its face and turns the image into a reading in the cloud. That way older utility meters count toward the water or energy balance without replacement. See pulse counters.

Yes. Aqueduct data can be exported and integrated with your analytics platform, data lake or AI tools, and portfolio views export to CSV. Each reading keeps its device, zone, floor and site, so exported data arrives with its context rather than as bare tags. See integrations with Aqueduct.

Aqueduct verifies a saving by comparing metered use after a change with an adjusted baseline built before it. Below the baseline is a saving and above it is waste, so a hot month cannot hide waste or fake a result. Savings on controlled equipment use the same method, baselining that equipment's power first. See measurement and verification.

A weather-normalized baseline is the use a site should have had given the actual weather. Aqueduct builds it by regression on six to twelve months of history against cooling and heating degree days for the site's own location, then reports actual use against it, line by line. See weather-normalized baselining.

Yes. Upload six to twelve months of utility bills and Aqueduct builds the baseline from them, so performance results start on the first day instead of after months of monitoring. The bills also carry the rates, which turn the gap between expected and actual use into money. See the building energy baseline guide.

Yes. Aqueduct's whole-facility baseline follows IPMVP Option C, sub-metered boundaries such as a chiller feeder match Option B, and the regression follows the approach in ASHRAE Guideline 14. Aqueduct does not certify savings: where a program requires sign-off, a third-party verifier reviews the data. See IPMVP explained.

Aqueduct runs in 275+ sites across 21 countries, including 20M+ sq ft of commercial real estate. Deployments include 22 hospitals and 9 clinics in one of India's top hospital chains, a top 25 US airport and major dairy operations. Each one is described in the case studies.

Milvian AI holds ISO/IEC 27001, ISO 9001:2015 and SOC 2 Type 2, and is an AWS Advanced Tier Partner. ISO/IEC 27001 covers information security and ISO 9001:2015 covers quality management. These audits cover how the company runs; what Aqueduct does to equipment is covered by each device's command history. See trust and security.

Aqueduct stores data in the cloud, in the customer's region, so operational readings stay within the jurisdiction the customer chooses. For operators whose data must never leave their own network, an on-premises deployment runs the same products on local servers.

No. Operators and their consultants file; Aqueduct supplies the numbers. Metered energy, water and air data feeds the inputs that ENERGY STAR Portfolio Manager, GRESB, CDP, LEED and WELL ask for, so a report is compiled from measured values instead of estimates. See ESG and sustainability reporting.

aQ Assure is the Aqueduct product that keeps compliance evidence. It holds the continuous record a standard asks for, such as room conditions under ASHRAE 170 or water management checks under ASHRAE 188, and shows which readings fell outside limits and when. It supplies evidence; it does not make a building compliant. See compliance.

Start with a demo or a free site survey. A typical first step is one site and one problem, such as water loss or HVAC energy: share six to twelve months of bills and a meter list, and the survey returns a baseline and the first findings from existing meters. Book a demo or site survey.

It depends on the site, but a top 25 US airport went live on Aqueduct in two weeks with 5,000 meter points. Reading existing meters first and commissioning each new device as it is fitted keeps the schedule short. Read the airport case study.

Installation partners do. Milvian AI builds the software and does not install hardware itself. Partners fit meters, sensors and controllers, and technicians commission each device in aQ Atria, which tags it to its site, floor and zone and checks it reports before the job closes. See aQ Atria.

Yes. System integrators, MEP firms, installers and facility services firms can specify Aqueduct, install around it and deliver it to their clients, in existing sites and new builds. Milvian AI supports partners with platform guidance during survey and solution design. See the partner program.

Yes. Architects and MEP engineers can write Aqueduct into the specification at design stage, so the metering plan, controls points and data model are right before handover. Aqueduct is device agnostic, so the specification does not lock the project to one hardware vendor. See building Aqueduct in.