Energy and natural resources companies are implementing AI pilot projects, but should focus on a few end‑to‑end domains where AI can have an impact, such as asset and process performance, customer and technical services enhancements, or supply chain optimisation, says management consulting firm Bain & Co.

A survey of 800 executives in the industry showed that 68% expect AI to have a substantial or transformative effect on business performance over the next five to ten years.

However, fewer than 20% of those surveyed reported that their companies are scaling up AI applications with measurable impact or executing company-wide transformations on current operations.

Rather than spreading effort across dozens of pilots, companies can start by selecting two or three end‑to‑end domains where AI can address material profit and loss or system challenges, the firm says.

Within each domain, the most successful organisations define their initial AI use cases with clear outcome metrics and ownership, thereby directly tackling the gap between high expectations and unclear value, says Bain & Co.

Further, progress beyond small‑scale experimenting is broad but shallow. Oil and gas and chemicals executives report the most headway so far. Customer service, research and development and operations and maintenance are currently the most mature applications.

For example, a multinational upstream oil and gas operator ran an AI‑enabled diagnostic to determine how a mature offshore asset could reach its full potential. It used data science tools and large language models to clean up and connect previously unlinked data sets, which helped uncover opportunities in asset care, logistics and well performance economics.

AI helped the company better forecast each well’s production, develop a marine logistics data model and create an equipment maintenance database directly linked to associated work orders and deferments.

The company made actionable plans for more than a dozen performance improvement initiatives valued at more than $15-million a year, with clear timelines, decision points and initiative owners.

Meanwhile, the energy and natural resources executives identified organisational, data and operating models as challenges. Nearly half of executives say the biggest impediment to AI achieving the desired outcomes is that the outcomes are unclear or have no link to business value, Bain & Co points out.

Additionally, shortages of technical expertise and poor data availability and governance were also identified as bottlenecks.

“The distance between AI's promise in energy and natural resources and its impact today is real, but it is not a technology gap. The obstacles that matter most are organisational and structural, including project outcomes that are unclear or untethered to business value, scarce talent, immature data and governance and operating models built for a pre-AI era.

“These are problems executives can act on directly, and the companies establishing an early lead in AI are focusing on redesigning processes around AI, establishing sufficient initial foundations, and developing a repeatable playbook for scaling-up impact,” Bain & Co says.

OPERATIONAL REDESIGN

Bain & Co recommends that energy and natural resources companies redesign processes and roles and apply AI where it adds business value, as many AI initiatives try to bolt models onto existing workflows. However, moving from pilots to production requires updating the function itself, such as maintenance planning or grid operations, and not only updating the tools.

The most effective companies start by mapping the current process from start to finish and identifying decision points, handovers and bottlenecks. This helps to guide the redesign of roles, metrics and ways of working around embedded AI, which might involve planners, dispatchers, technicians, traders or schedulers using AI‑enabled recommendations as part of standard routines, says Bain & Co.

Early adopters are also beginning to integrate AI agents into certain processes while keeping humans in the loop. The final step includes retiring legacy reports, tools and approvals that would otherwise drag AI‑enabled processes back to older operating models, it says.

The consulting firm also recommends that companies build sufficient foundations in data, technology architecture and talent for key domains.

Companies that report success focused first on the data sets, governance rules and platform capabilities needed to industrialise AI in priority domains.

For utilities, this could be asset and outage data for grid optimisation, while oil and gas companies could focus on management and sensor data for operations and maintenance to industrial AI.

In agribusinesses, integrated spending and contract data for procurement could serve as priority domains to industrialise AI.

A multinational agribusiness focused its first wave of AI investments on non‑commodity procurement and prioritised seven AI tools.

To build a solid initial foundation, it cleaned and restructured its spending data and embedded generative AI‑enabled classifiers, contract optimisers and negotiation assistants directly into the company’s category management workflows.

The result was higher savings targets, faster strategy development, materially better compliance and reporting and a foundation for later AI projects in manufacturing and logistics, reports Bain & Co.

Further, companies reporting successful AI projects establish small, cross‑functional teams that combine domain experts, data engineers and AI specialists. They use early wins to build capabilities and confidence and to guide subsequent investments in common platforms and data products.

This methodical approach will be much more effective than launching broad, abstract data lakes or generative AI platform initiatives, Bain & Co says.

Companies that move beyond pilot mode tend to align their scaling-up efforts with their internal culture and regulatory context.

This typically involved deploying an AI operating model that clarifies who owns value, who builds and runs solutions, and how funding works. This could be a small, central AI team in addition to empowered business‑unit squads in priority domains.

Additionally, companies reporting beneficial impact from AI implement lightweight but effective governance processes that prioritise, risk and responsible AI, which is critical in regulated energy and natural resources sectors.

Similarly, they develop a playbook for going from idea to scaled solution, with common patterns for discovery, design, pilot, industrialisation and change management, the consulting firm says.

For example, a European utility company found that its AI experiments were scattered across business units with little central visibility. It designed a new AI operating model informed by a deep dive into its processes, governance, skills and technology, as well as a scan of external benchmarks.

The new approach included creating an AI centre of excellence and responsible AI committee. The company also developed clearer protocols for managing AI costs, a technology architecture blueprint that generative AI tools could easily work with, and a detailed rollout plan to achieve AI at scale within six to nine months, Bain & Co says.

Leaders must decide how much responsibility to concentrate in a central AI group or in empowered business-unit squads; how to fund, govern and sequence investments so early wins compound rather than fragment; and where to build proprietary intelligence rather than buy or partner.

The bottlenecks hampering energy and natural resources companies from realising value from AI are not intractable. The executives who confront them early, with clarity on where AI creates value and the discipline to scale it, have a chance to do more than close the gap between expectation and impact, Bain & Co says.