Overview
Over the past decade, oil and gas operations have become very good at collecting data. A modern AVEVA PI System captures millions of values a day — pressures, flows, temperatures, runtimes — across thousands of measurement points. The instrumentation works, the data lands, and the archive keeps growing. The harder problem now is not capturing the data; it is getting a reliable answer out of it quickly enough to matter.
At the same time, the conditions around that data have shifted. Plant knowledge is spread across people, spreadsheets, and tribal memory rather than written down in one place. The experienced engineers who know exactly which measurement holds which value are stretched thin or are retiring, taking hard-won context with them. Operations teams are asked to make faster decisions from more complex systems with fewer of the people who understand them best.
This article looks at how MetaFactor’s IntelX closes that gap. IntelX is an AI-enabled system that lets people ask questions about live operational data in plain language and get back direct, grounded answers. What makes it usable in an operational setting is where the AI sits: at the edges, to understand the question and to phrase the answer — never in the middle, doing the math.
Key Questions
When teams look at how much time goes into pulling answers out of operational data, the questions that come up most often are these:
- Why does answering a simple question still take so long when we collect this much data?
- When I get a number, how do I know it came from the right asset and the right reading?
- How do we get answers to the people who need them without leaning on the handful who know the system?
These are not really technology questions. They are about speed, trust, and access — the difference between data that sits in an archive and data that supports a decision in the moment.
How IntelX Works
A question goes in as plain language and a grounded answer comes back. In between, the work is split deliberately. AI interprets the question — what is being asked, about which asset, over what period. Purpose-built software then does the reliable part: it retrieves the data from the connected sources, resolves the correct asset and measurement, and computes the result through an approved definition. AI phrases the final answer in plain language and shows the values it was built from. The arithmetic never runs inside the AI, so the same question always produces the same answer.
One System, Many Sources — Works Where You Work
Retrieval — Reading the Right Value
The most common questions are also the most basic: what is a value right now, what was it at some point in the past, or how has it moved over time. Answering “what is the oil rate on well 12 right now” normally means knowing which cryptically named measurement holds the value, pulling it into a trend or a spreadsheet, and reading it off. Multiply that across a shift and across everyone who has a question, and a large amount of skilled time goes into retrieval rather than decisions.
IntelX removes the lookup. A person asks in plain English, and the system resolves the question to the correct asset and measurement before returning the current value, a historical value, or a full time series — each with its source attached. No measurement names, no query syntax, no spreadsheet.
Aggregation — From Readings to Operational Totals
Many answers are not a single reading but a roll-up. Total flow over the last twelve hours, average pressure across a shift, combined volume across every well on a pad — these aggregate values over time, across the asset hierarchy, or both. Built by hand, asset by asset, this is the kind of effort that does not scale and tends to get skipped under pressure.
IntelX works from the asset model rather than from hardwired logic, so the same approach handles all of it. Ask for a windowed total and it aggregates over the period; ask for a figure across a pad or a battery and it rolls the values up through the hierarchy. One consistent method, whatever the asset or the window.
Comparison — Measuring One Asset Against Another
Some of the most useful questions are not about one asset in isolation but about how it stacks up against its peers. Which well is underperforming, which pad is running hot, which unit is drifting from the rest — the answer is a comparison, and building it by hand means pulling the same figure for every asset and lining them up.
IntelX does that in one step. Ask “what is the cumulative steam-oil ratio on Pad 3, and how does it compare to the other pads,” and it computes the same metric across the peer set, rolls each one up through the hierarchy, and returns a ranked result. One question, every asset in the group, no manual assembly.
KPIs and Compliance — Answers You Can Defend
The questions that carry the most weight are rarely raw readings. They evaluate a defined metric or check a value against a limit — and a wrong or invented answer here is costly, because it ends up in a report, a decision, or a regulatory submission. This is where the split between AI and software matters most, and where IntelX is most deliberate.
It evaluates a KPI through an expert-approved formula — “what is our sulphur recovery efficiency, and are we above our licensed minimum” — and returns the figure together with the readings it was built from, so the result can be traced back to its inputs. It also checks a value against a regulatory or operational threshold — “are we within our Directive 060 venting limit this month” — and returns a clear status with the directive cited. The AI never selects the asset and never does the arithmetic, so the same inputs always produce the same result. An answer you can put in front of a regulator is worth far more than one that is merely fast.
Reliable as You Add More
Every answer is only as good as the definition behind it, and this is where hand-built solutions tend to break down. Spreadsheets degrade as assets are added, and the reasoning behind a given calculation slowly turns into folklore that no one can fully account for.
IntelX addresses this at the point where a metric is defined, before it is ever used to answer a question. Definitions are sourced from your own standards — the directives, manuals, and references your experts already rely on — validated against known inputs and expected results, and reviewed and signed off by a subject-matter expert before they go live. Each one carries a record of where it came from, so its origin is auditable rather than assumed. The expert confirms the meaning; the software guarantees the arithmetic. Correctness is built in at definition time, so the catalogue of questions can grow without eroding the trust in the answers already in it.
We Can Help
IntelX sits on top of the systems you already run — it adds the answer layer above your operational data without replacing anything beneath it. Since it is built to an open standard, it is not tied to one interface: IntelX can be used anywhere, inside any AI or software tool your teams already work in — a chat assistant, a dashboard, an analytics environment, or your own applications.
MetaFactor builds and supports the operational data foundation that answers like these depend on. As Calgary AVEVA PI and AVEVA PI AF experts, we work with teams who want their people to get more out of the data they already collect, with less effort and more confidence in the result. If you would like to talk through how your teams use their operational data — and where something like IntelX might fit — get in touch.