In a discussion on digital innovation for offshore operations, the approach centers on a practical, problem-first application of technology, where artificial intelligence is used only when it adds clear value alongside traditional engineering.

Remote Operations as a Catalyst

The Covid-19 pandemic forced a rapid shift to remote monitoring and control of wells, which became a permanent structural change. This shift reduced personnel transport and reshaped offshore economics, leading to significant time savings, such as 15 well days saved on certain drilling campaigns in Argentina. This evolution formed the foundation for Baker Hughes' digital platform for autonomous well construction, Kantori, which is now essential for economically viable operations in harsher, deeper offshore environments.

Advancing Autonomous Drilling

Autonomous drilling is progressing quickly, though still in early stages. The company regularly achieves 15-20% efficiency gains, with some wells performing even better. It has successfully executed completely autonomous steering sections in Norway, resulting in faster and better wells. Specific, time-consuming tasks like pipe tripping, which accounts for about 16% of drilling time, are natural candidates for automation. The greatest potential for efficiency leaps is seen in international unconventional plays, where new entrants can bypass earlier learning curves.

A Hybrid, Engineering-First AI Philosophy

The strategy does not pursue AI for its own sake but starts with specific operational problems. AI is deployed only when it adds value, almost always in tandem with established engineering models. A key example is Electrical Submersible Pump (ESP) failure prediction: using AI alone yielded 30-35% accuracy, but combining a physics-based fluid movement model with machine learning boosted accuracy to around 70%. This hybrid approach was the breakthrough.

Layered AI Applications and Tangible Results

Baker Hughes now applies AI in several layers: physics-plus-AI models, AI reservoir decline forecasting, field optimisation algorithms, and systems monitoring (and sometimes executing) operational adjustments. The operational gains are significant, exemplified by a Middle Eastern project with 200 wells that saw a 20% productivity increase alongside an 18% reduction in power use. For mature fields, optimisation through ESP tuning, gas plunger control, gas lift adjustment, and production AI models delivers incremental but cumulative gains, with gas plunger optimisation alone adding 3-4% in marginal fields.

Future Priority: Trust and Optimization at Scale

The top future priority for digital improvement in upstream operations is field optimisation at scale, crucial for reducing cost per barrel amid margin pressures. While AI and automation are central tools, the biggest adoption barrier is not technology but trust. Operators will not use solutions they cannot understand; there is no tolerance for "black boxes." Even with proven solutions, adoption requires that operators and engineers retain ultimate control, with AI serving strictly as a decision support tool, not the top of the decision tree.