Agentic AI in energy:  Keys to unlocking value

Enterprises across a wide range of industries are investing in agentic AI to drive new levels of efficiency by automating complex tasks and enabling innovation. According to Gartner, 61% of businesses have already invested in agentic AI.¹ However, its research also reveals that some may struggle to realize the full value of their investments, forecasting that over 40% of agentic projects will be canceled by the end of 2027.²

John Chu, Vice President Global Energy at EPAM, explains: “The promise of agentic will only be realized if the industry can achieve four key objectives: building trust, ensuring flexibility, enabling interoperability, and providing a natural user experience for agent-human interactions.”

Building a foundation of trust

Trust provides the underpinning of success. As John Chu explains: “Just as we don’t entrust decisions to individuals we can’t hold accountable or don’t know, we should be equally cautious about delegating decision-making to agents without transparency regarding their origin, developers, training methods, underlying data, and the reasoning behind any particular recommendation.”

That makes provenance (who built or certified the agent), audit trails showing actions and accountability, “explainability” to make reasoning visible, and other measures of social validation crucial.

“Imagine a drilling-plan agent that produces a design but explains which offset wells it used, the risks it flagged, the NPT events those risks were based on, the trade-offs it weighed,” adds John Chu. “Or a reservoir simulation agent that clarifies the history match quality, the key subsurface uncertainties, and alternative scenarios before recommending a development strategy. Transparency like that is how agents earn the same trust we extend to people.”

Embed flexibility into your technology

Vendor lock-in is a real and perennial fear with adopting any technology. No organization wants to stake its future on a single provider or platform, especially with technology that is evolving as fast as agentic AI.

However, slowing innovation because of that fear means missing opportunities. At this stage of the agentic paradigm, leaders should press ahead while embedding flexibility.

“CIOs should consider portable architectures that separate logic from infrastructure, data stored in open formats that can move if needed, and standards that let agents run across environments,” says John Chu. “Think of it as building an exit strategy from day one so experimentation doesn’t mean entrapment.”

Focus on interoperability

Energy workflows already span dozens of products and platforms, both legacy and modern. Agents will only add to the mix.

According to John Chu, the goal shouldn’t be to replace these systems but to connect them. Interoperability means that agents should have the ability to communicate with one another in natural language without custom integrations between the applications themselves.

Achieving this type of interoperability will require open standards; shared registries where agents are certified and trusted; and governance agreements so operators, service companies, and regulators can collaborate.

Reimagine the user experience

One of biggest shifts we anticipate agentic AI to trigger is with how users interact with software. Today, modern applications are rigid and opaque. AI agents will demand a new model where humans and agents engage together in a shared workspace, with legacy platforms operating in the background.

John Chu explains: “An agentic user experience could include a common canvas where humans and agents collaborate, natural language goals instead of manual steps, and ‘explainability’ panels showing why a decision was made and which agent made it.”

Over time, humans will set the direction and guardrails, while agents handle execution and adapting to new conditions.

John Chu concludes: “That’s the paradigm shift. The future of software in energy isn’t forms and clicks – it is people and agents jointly collaborating in real time.”

AI service providers like EPAM help to manage the upkeep, tuning, and continuous training of agentic solutions to ensure accuracy and trust over time. As an unbiased and independent engineering partner, EPAM also plays a critical role in helping businesses to select the right technologies while maintaining the flexibility they need.

^{1 Gartner, “Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027” June 2025}

^{2 Ibid}