An AI assistant for facilities teams that answers with the evidence attached.
aQ Assistant is a multi-agent AI assistant built into Aqueduct: ask about your sites, zones or devices in plain language and it queries the platform data to answer. It knows which site and page you are looking at, returns analysis rather than a single number, and warns you when the data behind the answer is too thin to trust.
Ask about any site in plain language and get the answer with its evidence.
Ask about sites, zones and devices without knowing the hardware
- No meter hunting
Ask how a site compares to last week instead of walking from the main meter down to each sensor.
- Knows your context
The assistant takes the site and page you are looking at as context, so questions stay short.
- Suggested prompts
Starting questions sit in the side panel, so a new team member gets a useful answer on day one.
Answers that show consumption, demand and the trend behind them
- Week by week
Answers compare this week with recent weeks in a table, so the change and its timing are clear.
- Projected month
Consumption to date is projected to a full-month equivalent, so the team sees where the month will land.
- Peak and average
Average and peak demand come with the answer, so a load problem is visible without another query.
A warning when the data is too thin to trust
- Coverage checked
The assistant reports how much of the period the data covers before drawing a conclusion from it.
- Weak days flagged
Days with low coverage are named, so a week-over-week comparison is not trusted when it should not be.
- Points to the device
Device health shows which sensor caused the gap, so the fix goes to the right place.
Hunting for answers versus asking for them
- A usage question means knowing which meter feeds which floor.
- Answers come back as one number with no context.
- Nobody says when the data behind a number is incomplete.
- Comparing weeks takes an export and a spreadsheet.
- Only the controls specialist can answer the question.
The Aqueduct way- Ask in plain language, and the assistant finds the right meters.
- Answers include trend, peak demand and a projected month.
- A data quality warning comes with every thin answer.
- Week by week comparisons arrive in the answer itself.
- Anyone on the facilities team can ask and get evidence.
The question is simple. Finding the answer is not.
Without an assistant, a usage question means walking the data from the main meter down to individual sensors.
You need to know the hardware
Answering 'how much did this site use this week' usually means knowing which meters feed which panel. Most people asking don't.
Answers arrive without caveats
A number from a week with missing readings looks as solid as one from a full week. Nobody tells you which is which.
Context gets lost
Every question starts from zero: which site, which period, which unit. The tool does not know where you are.
How aQ Assistant answers
It sits in a side panel next to the page you are on.
Ask
Pick a suggested prompt or type your own: 'Ask about your sites, zones, or devices.'
Take context
The assistant knows which site and page you have open and uses that as the starting point.
Query the data
Multiple agents use tools to query the Aqueduct database in aQ Core, the same store every outcome reads from.
Answer with evidence
You get the figures, a comparison, key observations and, where it applies, a data quality warning.
What a real answer contains
Asked for an electricity consumption summary for a site and how this week compares to recent weeks, a live answer included all of the following.
Consumption to date
What the site has used so far in the period.
Coverage of the month
How much of the month the readings actually cover.
Projected full month
What the month looks like if the pattern holds.
Average and peak demand
Both, because the peak is often what the tariff charges for.
Week by week
A comparison table across recent weeks, plus key observations in plain language.
Data quality warning
A flag that certain days had very low coverage, so the week-over-week comparison was unreliable.
Why the data quality warning matters
An assistant that always sounds confident is a liability in a plant room. If a sensor dropped off for three days, a weekly comparison built on those days is wrong, and the person reading it has no way to know.
aQ Assistant reports how much of the period the readings actually cover. When the readings behind an answer are thin, it says so, flags the days affected, and tells you the comparison is not reliable. That is the difference between an answer and evidence.
An answer without its evidence is just another opinion about the building.
Talk to our teamQuestions each seat can ask
“Give me an electricity consumption summary for this site. How does this week compare to the last few weeks?”
Consumption, coverage, projection, average and peak demand, a week-by-week table and observations, with a warning if the data is thin.
“Which site, zone or asset changed this week?”
The assistant answers from the portfolio data in aQ Core, with the evidence attached so you can open the source.
“How much of this month do the readings actually cover?”
Coverage is part of the answer, with a data quality warning when some days are too thin to compare.
Where it saves time
The assistant does not replace the views. It gets you to the right one faster.
A usage question needs someone who knows the metering layout
Ask in plain language on the site page; the assistant finds the meters
Consumption, peak demand and trends without knowing the hardware
A weekly comparison hides gaps in the readings
Answers carry coverage and a data quality warning
Decisions made on numbers you can defend
Every report starts from a blank page
Context from the page you are on sets the site and scope
Faster first answer, fewer wrong assumptions
What it is, and what it is not
AI is the method, not the product.
It reads, it explains
The assistant queries platform data and explains what it finds, in plain language.
It is not the control path
Equipment control runs through Smart Controls, only where enabled and with owner approval.
It can stay on site
For data that cannot leave, Aqueduct runs on-prem AI on NVIDIA infrastructure.
Where teams put it to work
A weekly summary in one question
Ask for an electricity summary for a site and how this week compares, and get consumption, peak demand and observations to act on.
Learn more
Which sites need attention this week
Ask across the portfolio which sites are running above their baseline, then open the site that matters and see why it moved.
Learn more
The assistant can stay on site
Where operational data cannot leave, the assistant runs with the platform on premises on NVIDIA infrastructure, with the same answers and evidence.
Learn moreRelated
Common questions
A multi-agent AI assistant inside Aqueduct. You ask about your sites, zones or devices in plain language, and it uses tools to query the platform data and answer with the evidence attached.
Yes. It is context aware: it takes the site and page you have open as the starting point for the question.
It tells you. A live answer flagged that certain days had very low coverage, so the week-over-week comparison was unreliable.
The assistant answers questions. Equipment control runs through Smart Controls, only where control is enabled and with owner approval.
Where data cannot leave, Aqueduct deploys on-prem AI on NVIDIA infrastructure. The scope of an on-prem deployment is set during scoping.
See Aqueduct on your own facility.
Bring a site, a utility bill or a question. We will show you where the loss is.