Forward deployed engineers in oil and gas: getting AI past the pilot and onto the shift
Why industrial AI in oil and gas stalls after the pilot, and how Voho's forward deployed engineers get it onto the shift: read-only to the control layer, connected to PI, SAP PM and the document system, evaluated on your own past events, with engineers signing off every action.
Oil and gas has run more AI pilots than almost any other industry, and scaled fewer of them than it should have. The pattern is familiar: a data science team builds a model on an extract, it performs well on the slide, and it never becomes something an engineer opens on shift. The model was rarely the problem. The distance between the model and the operation was.
Forward deployed engineering closes that distance. Voho's engineers work from the operator's or contractor's own site, connect the AI to the systems the shift already uses, and stay until it is part of how the work is done.
Why industrial AI stalls after the pilot
- The pilot ran on an extract. Production needs a governed connection to the historian, SAP PM and the document system, reviewed by OT security.
- Nobody replayed it against the plant's own failures, so engineers have no reason to trust it.
- The output lands in a new dashboard nobody on shift opens, instead of the notification or work order they already use.
- It was built by people who have not stood on a rig floor or in a control room, and it asks questions a crew would never ask.
What the FDE does on an industrial site
| Step | What the engineer does |
|---|---|
| Pick one asset or workflow | A compressor train, a drilling campaign, a set of wells: somewhere the value is easy to measure and the data already exists. |
| Connect the harness | Read-only connectors to the historian, SAP PM or Maximo, WITSML and the document system, inside your network, reviewed by your OT security team. |
| Evaluate against past events | Replay the last failures and near-misses. The agents are measured on whether they would have caught them, and how noisily. |
| Go live on shift | Engineers use it with approvals on. Notifications and work orders are drafted in the system that owns them and committed only after sign-off. |
| Widen | The next asset class reuses the same connectors, permissions and evaluations, so it is configuration, not a new project. |
What gets automated across the value chain
| Segment | Typical first automations |
|---|---|
| Upstream | Daily drilling reports drafted from rig data; drilling advisory on WITSML streams; ESP and artificial-lift health, well by well; well files answerable in one question |
| Midstream | Compressor and pump health against the operating envelope; pressure and flow deviations traced to the segment; shift handovers written from the alarm log |
| Downstream | Bad-actor lists from DCS alarms and SAP PM failure codes; turnaround scope assembled from work orders and inspections; procedures and MOC records searchable with citations |
The lines an FDE does not cross
- Read-only to the control layer. Voho reads SCADA, DCS and historian data and never writes a setpoint or touches a safety system.
- Inside your network. The harness and the models run in a Saudi region or on your own servers, including air-gapped sites.
- Permissions you already have. Each agent inherits the access of the person using it.
- Engineers sign off. Nothing is raised, closed or scheduled until a named engineer approves it.
- Every step logged: query, tool call, model output and approval, streamable to your SIEM.
Who goes on site
Voho's oil and gas work is led by people who have worked the industry: Aamir Shahzad, with thirty years in drilling operations, rig management and well services across Saudi Arabia, the UAE, Pakistan and North Iraq, and Imtiaz Ahmed, who spent most of his career at Arabian Drilling Company. They are the reason our engineers ask the questions a rig crew or a reliability engineer would actually ask. Our delivery team is already on site at Aramco in the Eastern Province, taking an AI call centre live, and the same forward deployed model is how we take industrial work onto the shift.
If you have one asset and a year of events, that is enough to start. Book a working session, or read the oil and gas page for the harness in detail.
Frequently asked
- Why do AI pilots in oil and gas fail to scale?
- Usually not because of the model. Pilots run on data extracts rather than governed connections, are not evaluated against the plant's own past failures, deliver into a new dashboard instead of the systems the shift already uses, and are built without people who have worked the operation.
- Does Voho's AI write to SCADA or DCS?
- No. Voho is read-only to the control and safety layer. Actions go only to business systems such as SAP PM or Maximo, and only after an engineer approves.
- What does a first oil and gas AI project look like?
- One asset class or one workflow, connected read-only inside your network, evaluated against your past events, then used on shift with approvals on. Later use cases reuse the same connectors, permissions and evaluations.
Keep reading