How TotalEnergies Gas Mobility is Scaling Hydrogen and BioCNG Refueling Operations with Connected Data and Agentic AI
Why Energy Operations Demand Scalable, Always‑On Expertise
Energy operations don’t get to pause. Assets run across multiple geographies, teams rotate, and critical knowledge is often concentrated in a limited set of practices, playbooks, and diagnostic patterns that must be applied consistently—site after site, shift after shift. That reality is exactly why many energy companies are investing in connected operations technology and exploring Agentic AI not as a novelty, but as a practical way to optimize execution, shorten response times, and scale operational capability globally.
At Siemens, we partner with organizations to transform industrial data into actionable insights at scale. One example is our collaboration with TotalEnergies Gas Mobility, the team is strengthening day-to-day operations and maintenance of hydrogen and bioCNG refueling stations through connected data, while also evaluating Agentic AI to make operational support faster, more consistent, and easier to scale worldwide.
The Hidden Constraint of Outsourced Operations: Scarce Expertise at Scale
In many energy environments, operations and maintenance are frequently outsourced. That model can be effective, but it introduces a familiar constraint: specialized operational expertise becomes a scarce, high-demand capability. Troubleshooting paths, failure signatures, and “what to check first” logic must be applied across a large and growing asset base—often with variation by site, contractor, and local practice. The challenge is less about who is available in the moment and more about how expertise is captured, standardized, and embedded into repeatable workflows so that decisions remain consistent, auditable, and scalable across regions.
From Reactive Troubleshooting to Structured Operational Management
This is where connected operations technology becomes foundational. When asset data is consistently captured and contextualized, teams can move from reactive troubleshooting to structured operational management. The goal is not “more data” or “more dashboards.” The goal is to enhance decision-making in real-world conditions, enabling teams to detect anomalies earlier, troubleshoot with improved context, and align on a consistent response approach, even when third-party partners and distributed teams are involved.
Using Agentic AI to Codify and Scale Tacit Operational Knowledge in Hydrogen and BioCNG Refueling Operations
Agentic AI builds on that foundation by addressing the most challenging aspect to scale: tacit, experience-based expertise. The most valuable operational knowledge often resides in patterns—recognizing early indicators, separating noise from signal, and selecting the fastest path to the root cause. That’s difficult to document and even more challenging to implement across different geographies. Agentic AI offers a practical way forward by translating proven diagnostic approaches into repeatable “skills” that guide teams with the same rigor and consistency every time.
We’re building a way to codify operational know-how—so proven diagnostic and decision practices become reusable capabilities we can scale across our hydrogen and bioCNG refueling stations worldwide.”
David Dowling, Manager of Data and Connectivity, TotalEnergies Gas Mobility
Engineering AI Agents for Repeatable, Real‑World Performance
A key learning from their early work is that strong agents are engineered, not improvised. Teams often assume the value comes primarily from the AI model, but in operational settings, the differentiator is how clearly the work is defined—inputs, constraints, and success criteria.
We approach agent design like engineering: define the outcome, specify the inputs and constraints, and write instructions that are precise enough to perform reliably in real operations.”
David Dowling, Manager of Data and Connectivity, TotalEnergies Gas Mobility
That mindset matters because operations require repeatability. When an agent is designed with explicit inputs, clear constraints, and a consistent output format—and then tested against real scenarios—you’re not building a general-purpose assistant. You’re building an operational capability that behaves predictably, which is what teams need when uptime, safety, and performance are on the line.
TotalEnergies Gas Mobility is also grounding this work in measurable outcomes. One insight they’re seeing early is that quality is emerging as a leading driver of low OEE, which immediately shapes where to focus. Instead of building broad, generic AI support, the emphasis shifts to targeted solutions: identifying what drives quality loss, detecting signals earlier, and guiding corrective action in a way that reduces recurrence over time. This is a strong blueprint for applying AI in industrial environments—start with a high-impact driver, build a repeatable workflow, and continuously refine it based on operational feedback.
Looking ahead, the most significant shift is moving from reactive support to proactive operations. Many organizations start with a model where someone identifies a problem and then asks the AI for assistance. That can help, but it still depends on issues being identified quickly and accurately described. The more effective direction is when the system detects the issue and prompts the user with context and recommended actions—essentially: you have a problem, here’s what’s happening, and here’s how to respond. Done well, that improves speed, consistency, and accuracy while reducing the operational drag created by scarce expertise.
Learn more about Siemens Insights Hub
