A strategic note for oil & gas service providers
By Sreetaran Ponnuthurai · Helios Energy Advisory · September 2026
Data quality is the foundation of any digital or AI initiative. Here is what that takes in oil & gas — and how to build it.
Why it matters
In close to two decades working in oil & gas projects and operations, I have found the pattern to be consistent: technology without the right foundation will not deliver value effectively. And when it comes to AI, that foundation is data. An AI model can only work with what its data teaches it, and when that data is inconsistent, incomplete or contradictory, the model does not tell you so — it simply gives you an answer that looks reasonable and cannot be trusted.
In oil & gas this matters more than in most industries. Our decisions carry real safety, environmental and financial consequences, and they increasingly rest on data drawn from our own operations. So data quality is not a technical housekeeping task to be delegated. It is a core risk to be owned and governed — and the industry evidence bears that out:
80%+ of AI projects fail — around twice the rate of ordinary IT projects (RAND, 2024).
~3% of energy firms describe their AI use as advanced or highly integrated (DNV, 2024).
Where data quality tends to break down
None of these are exotic failures. They are the ordinary friction that builds up over years of operations, systems and hand-overs — and each one quietly undermines whatever you try to build on top of it. In my experience the most common are these:
The same equipment named differently across assets, systems and vendors — so no two datasets agree on what a thing is.
Sensor tags that drift, and units that differ between systems — the same reading arrives in three shapes.
"Completed" that means different things on different platforms — status without a shared definition.
The reason behind a decision never recorded as data — critical knowledge that leaves when people do.
History held in PDFs, spreadsheets and handwritten logs that systems cannot read.
Four steps to a data foundation you can trust
Building that foundation is deliberate work, not a one-off clean-up. Four steps do most of it:
One data model and naming standard. A single, agreed name for each item of equipment, tag and event — across every asset and system.
Capture quality at the source. Validation, controlled fields and required entries at the point of entry, rather than cleaning data after the fact.
Measure data quality. Track completeness, accuracy, consistency and timeliness for the datasets that matter most.
Assign clear ownership. A named data steward, accountable for each critical dataset — not a shared responsibility that belongs to no one.
Sustaining data quality is a change, not a tool
Good data is created by everyday actions right across the organisation. No tool sustains that on its own; it asks people to work differently, and that has to be led. Three things keep it alive:
Senior ownership. Accountability for data quality held at a senior level — not left unassigned or pushed to IT.
Feedback to the frontline. Show the people who capture data how it is used downstream, so its value is visible where it is created.
Aligned roles and incentives. Make stewardship part of the role, and recognise it accordingly.
Digital transformation follows an order
AI and analytics repay the investment only when the layers beneath them are in place first. Skip one and the initiative rarely fails outright — it just quietly underdelivers, and the model takes the blame the data earned. I find it helps to picture a foundation you build from the bedrock up:
Foundation — one data model, capture discipline, connected sources. Load-bearing; everything above rests here.
Governance — ownership, quality standards, and a way to measure them.
Culture — stewardship and incentives that sustain the practice.
AI & Analytics — applied on data that can be trusted.
How Helios helps
At Helios I work with oil & gas leadership across strategy, the data foundation and adoption — sequencing each initiative around the foundation it genuinely needs, rather than starting with the model and hoping the data holds. The aim is real, lasting capability, not box-ticking.
Set the strategy. A digital and AI roadmap sequenced around the data foundation each initiative requires.
Design the foundation and governance. The data model, ownership and quality standards that make AI outputs reliable.
Lead the change. The stewardship, incentives and communication needed to embed new practice on the frontline.
In summary
Data quality is a prerequisite for AI, not an afterthought. Access to advanced models is now broadly equal across the industry. The real differentiator is the quality of the data beneath them, and the discipline to maintain it. The encouraging part is that this can be built — in the right order, starting now.
I'd welcome a conversation about building the right foundation.
Sreetaran Ponnuthurai Founder & Principal Consultant, Helios Energy Advisory
