Issue 02 - The Data Issue

Every operator I talked to at the Global Energy Show eventually landed on the same complaint, usually a few minutes in. Not enough AI, not enough autonomy. Too much data in too many places, and no clean way to get at it. A well database here, a SCADA feed there, a shared drive full of PDFs nobody has opened in a year.

This issue is about three companies that read that complaint differently. Denodo leads because their answer is the most counterintuitive: stop moving the data at all. Data Squared and Ezops follow, each solving a version of the same problem from a different angle. One from the business user's chair, one from the field.

If you're catching up, Issue 01 covered AIQ, DataRobot, and Data2 on autonomous decision-making. This one is about what happens before autonomy is even possible: getting the data into a shape a system can use.

Every enterprise data problem eventually comes down to the same bad options. Copy the data somewhere new and hope it doesn't drift out of date. Or leave it scattered across fifteen, twenty, sometimes thirty different systems and ask people to stitch it together by hand. Denodo has spent 25 years building a third option, and in oil and gas, where real-time well and meter data can span just as many disconnected sources, that third option is starting to look less like a nice-to-have and more like the only sane way to run a dashboard.

Traditional tools like Informatica or SSIS work by extracting data, moving it to a staging area, transforming it, and loading it into a warehouse. That process is slow, it's expensive, and it leaves you with a copy of data that already existed somewhere else. Denodo skips the copy entirely. The platform extracts metadata from all of a company's source systems and uses it to build what they call a logical data product: a single access layer that knows where everything lives without ever relocating it.

When someone queries that data product, Denodo pushes the query out to the original sources in real time, pulls back only the rows that matter, and assembles the result in memory. If a table has a million rows and a use case needs a hundred of them, Denodo queries a hundred. Nothing else moves. That has a security benefit and a cost benefit at the same time: less data in transit, less compute spent processing rows nobody asked for.

That story is the real pitch. Denodo's security framework governs access down to the column level through role- and attribute-based tags, so a junior analyst and a VP can query the same data product and see different things by design. For heavily regulated operators, that turns a 90-day legal review into something closer to a formality, because there's no new copy of the data for legal to review in the first place.

The platform deploys on AWS, Azure, Google Cloud, or Alibaba Cloud, and on-premises for operators who need the data to stay put entirely. Energy customers already running it include VTTI, BHP, and Engie Mexico, the last of which uses it to power web services for hundreds of thousands of natural gas customers. Across more than 30 industries, Denodo customers report over 400% ROI with payback in under six months. After 25 years in the market, that's not a startup's pitch deck. It's a track record.

Denodo solves the data access problem from the infrastructure side. Data Squared solves a version of the same problem from the other end: the business user who doesn't have a technical background and shouldn't need one to get an answer.

The platform connects to a company's existing systems by API, whether that's AWS, Azure, or on-premises, and lets people ask questions in plain language the way they'd ask a colleague. Why is well 15 underperforming against the other hundred wells in the field? What do sales look like this quarter by region? The platform returns visualizations and a direct answer instead of a support ticket to IT. There's a governance layer built in too: administrators can see who has access to sensitive data and reassign policies without a separate compliance tool.

Data Squared is three years old, backed by ATCO Ventures and Bow River Solutions, and runs lean at 17 people. That size shows up in how they talk about fit: this isn't built for a two-person shop with a spreadsheet. It's built for mid-size to large operators generating enough data that a plain-language interface actually saves someone real time. Oil and gas is already their strongest vertical, and they're in active conversations with several large Canadian operators.

The company is also co-building Edgeworks with ATCO Ventures, a physical edge-computing container that keeps data processing local for remote sites. For a mine or a remote field operation, that means sensor data never has to cross an international border to get processed. It stays exactly where it was generated.

Ezops has been solving the field data problem since 2016, which puts them in an unusual position for a company now selling an AI product: they built the hard part first, and AI came later.

The problem they set out to fix is one anyone who's worked a field job will recognize. An inspection gets logged on paper. A form gets filled out and left in a truck. An equipment reading goes into someone's notebook because the signal's bad and the app won't sync. Ezops calls this the localization problem: data trapped in scattered, disconnected places, which means nobody in the office gets a real-time view of what's actually happening on site, and nobody in the field benefits from patterns the company has already seen elsewhere.

That's the case Ezops makes for itself. Eight years of field data and eight years of understanding how oil and gas operations actually run gives their AI something to work with that a newer platform doesn't have. It shows up in specific features: point a phone camera at a handwritten field note and Ezops converts it into a work order automatically. Give the platform access to an inbox and it can parse incoming emails into work orders the same way. Ask it a direct question about historical data, which location had an issue in 2018, for instance, and it answers instead of asking you to build a report.

Connectivity is the other problem Ezops was built around long before AI made it fashionable to mention. Bad signal doesn't take the system down. It keeps logging locally and picks back up the moment a connection returns, AI included. For downstream, midstream, service, and construction companies working sites where a strong signal is the exception and not the rule, that's less a feature than a baseline requirement.

Issue 03 turns to companies watching the physical world instead of the data behind it: Field Eagle, Essetrack, and LiveEO, on inspection, corrosion, and satellite monitoring. In your inbox in two weeks.

If someone forwarded you this issue, subscribe at themudlog.com to get The Mudlog in your inbox.