APPEC: Energy industry moving swiftly from AI experimentation to operational deployment

The energy industry is moving beyond artificial intelligence pilots and proof-of-concept projects as companies seek to integrate AI into daily workflows and deliver measurable business value, although significant challenges remain, speakers at APPEC said Sept. 10.

Rather than treating AI as a stand-alone innovation project, organizations are increasingly focused on applying the technology to practical challenges across trading, risk management, operations and decision-making. These include processing large volumes of structured and unstructured data, automating routine tasks, identifying market and operational insights and helping employees make faster, better-informed decisions, they added.

AI adoption is also expanding across the energy and shipping industries, helping companies reduce operating costs, improve productivity and lower emissions. Speakers and delegates at APPEC said its use is likely to grow particularly rapidly in upstream energy, where exploration and prospecting involve significant uncertainty.

"Deploying AI requires a more integrated approach across the value chain. Data centers, energy sources and the broader energy mix, as well as access to water, need to be considered together," Nick Sharma, head of upstream insights at S&P Global Energy CERA, told the conference. "These should not be treated as separate asset classes because their interdependence is critical to building and scaling AI infrastructure."

In the maritime sector, AI is emerging as a tool for analyzing safety and compliance data, optimizing voyage planning and port logistics and assessing weather-related and maintenance risks. However, a shortage of specialized skills, the complexity of integrating disparate data sources and concerns about trust and transparency remain significant barriers to wider adoption, speakers at APPEC said.

Supporting trading systems

Agentic AI is expected to augment rather than replace existing energy trading systems, said Animesh Agrawal, director at ClearOpx.

Rather than displacing energy trading and commodity risk management systems, the next stage of AI adoption could involve an intelligent layer operating above them. This layer would help companies connect and interpret information across their operations, including structured data held in trading systems and unstructured content such as emails, messages and documents, Agrawal said.

By bringing these disparate sources together, agentic AI could support information retrieval, workflow automation, trade analysis and decision-making, while allowing existing systems to remain the core platforms for managing transactions and risk, he added.

However, the industry is still developing reliable ways to measure the value generated by AI, said Aditya Kumar Aggarwal, an advisor in the commodities and energy industry. More standardized approaches are needed to assess AI returns, he added.

The benefits of AI are unlikely to be uniform across the energy industry, Aggarwal said. Value creation will depend on factors including a company's business model, risk profile and market segment. For some organizations, returns may come from improved operational efficiency and lower processing costs, while others may benefit from faster decision-making, stronger risk controls or better use of data.

As companies integrate AI into everyday workflows, speakers at APPEC stressed that human judgment will remain important even as the technology takes on more complex tasks. Organizations should retain human review and approval at critical stages to manage risk, maintain accountability and limit the potential impact of inaccurate or fabricated AI-generated outputs.

They also identified infrastructure as a key differentiator in the success of AI initiatives. Beyond selecting powerful AI models, companies will need to strengthen their data architecture, security capabilities, computing resources and technology platforms to support deployment at scale.

Implications on trading

Physical energy and commodity trading continues to rely on a fragile chain of emails, text messages, spreadsheets and productivity tools, creating opportunities for delays, errors and disputes to enter trades, Tasja Botha, CEO of VAKT Global, told APPEC.

While presenting at APPEC in a session titled "Frictionless Trading: Eliminating Operational Risk in Physical Markets," Botha said fragmented information and manual handoffs remain significant sources of operational risk. Operators often need to compare nomination data with counterparty inputs, bills of lading, receipts and certificates of analysis and quality, while also reentering information into spreadsheets and internal systems.

The resulting inefficiencies can delay invoicing and consume employees' time with administrative work.

"One oil trader needed about 10 switches between three systems, five email round-trips and three people just to resolve a single quantity discrepancy before invoicing," Botha said. "Applying the right approach, that same fix can drop to as few as eight manual actions, and the industry is working hard to get it lower still."

She identified four areas where companies could initially focus their efforts: automatically validating trade facts by converting unstructured recaps and contracts into structured data; digitizing trade confirmations; using agentic tools to manage scheduling and inventory for vessels, barges and pipelines; and standardizing actuals and reconciliation to automate invoice checks and reduce disputes.