In 2025 AI Won’t Just be a Tool
Wendy Laursen
February 13, 2025
Some expert thoughts on AI in 2025:
“In 2025, AI won’t just be a tool; it will be a collaborator.”
“In 2025, the most productive employees will work alongside AI agents.”
AI agents are an advance on large language models (LLMs). LLMs at present are popularly used as creative and conversational search engines. They can’t directly interact with external tools or data processing system or autonomously set up systems to monitor and collect ongoing data. AI agents can, and they can proactively track and collect data streams to refine their decision-making. They operate autonomously and can maintain long-term goals.
Despite their far-reaching potential, the latest issue of Maritime Reporter features an article by Heather Combs, CEO of Ripple Operations, explaining that the technology is already within reach and not just for tech experts. “The beauty of modern AI-driven tools lies in their simplicity,” she says. Many platforms, like Microsoft Copilot Studio, Google AI Studio and Amazon Bedrock studio, allow for the creation and deployment of AI agents.
For training, they can be set up to follow commands such as: “Match recent incidents with corresponding training programs and suggest improvements.”
“AI-driven agents aren’t science fiction,” Combs says. “They’re practical, accessible and ready to make a difference.”
ADNOC is already implementing agents into its offshore business. In January, ADNOC and AIQ announced the successful proof-of-concept trial of ENERGYai, an agentic AI system that integrates with a 70-billion-parameter LLM with over 50 years of ADNOC’s knowledge and petabytes of its proprietary data to drive optimization and efficiency across the company’s operations.
The agents demonstrated significant improvements in the pace and accuracy of upstream exploration through rapid, precise and detailed seismic survey analysis, alongside relevant, actionable insights to support production optimization at ADNOC’s existing wells. ENERGYai was able to deliver its insights in intuitive natural language, enabling engineers to interact with it effectively and easily.
The results of the trial delivered a 70% improvement in accuracy in major seismic interpretation aspects and significant improvements in advanced reservoir monitoring and anomaly detection.
A recent article provides a demonstration of how agents might work to support predictive maintenance. The example given relates to an aircraft:
A Boeing 787 aircraft is scheduled to fly from Vancouver to Seattle with a planned stopover in Montreal. It encounters turbulence during the first leg of the flight and a harder-than-usual landing. Upon arrival in Montreal, the AI agent uses sensors and flight data to analyze the impact of the flight, including increased vibration levels recorded by sensors in the landing gear. It then presents this information to a LLM which predicts a potential issue with the braking system and recommends a thorough inspection to the maintenance crew.
There is already a next step underway in the development of agentic AI – super agents where multiple AI agents work collaboratively to solve complex tasks.
A McKinsey report released last month year indicates that biggest barrier to scaling AI is not employees - who are ready - but leaders who are not steering fast enough.
Another expert thought on AI: “I’ve always thought of AI as the most profound technology humanity is working on . . . more profound than fire or electricity or anything that we’ve done in the past,” said Sundar Pichai, CEO of Alphabet.
