Three Strategies for Seamlessly Incorporating AI into the Modern Energy Equation

In the ongoing effort to improve the profitability of energy production and delivery while mitigating the risks, companies are incorporating artificial intelligence (AI) to drive more powerful, effective analytics. This technological leap is critical as the industry navigates evolving demands, embraces new sources, and strives for greater efficiency, reliability, and sustainability. However, some energy companies stumble when they begin their first AI initiative, often because their data foundations are simply not ready to provide their AI applications with the actionable results they need. Here, I will cover three crucial strategies for getting your data foundation “AI ready.”

Strategy 1: Enable Transparency: AI has the ability to sift through petabytes of data to extract powerful insights, many of which will sound impressive, assuming they are true. However, it is natural that someone will ask, “How can we prove that this is true? What data is the AI using to come up with this insight?” This question might be raised at an executive meeting, or it might be raised by any team member who is leveraging the AI for analysis. Regardless, it’s a question that should always have a ready answer, and with the right strategy in place, this can be greatly facilitated.

COMMENTARY

This strategy involves the ability to trace the lineage of all data-centric AI claims (and almost all such claims are based on data) all the way back to the original input data sources. In data management, this is a relatively straightforward process when one is working with a consolidated data source, such as a data lakehouse. However, it becomes much more complex when organizations are using AI, often in real-time, against multiple other sources in addition to a primary data lakehouse, such as on-premises systems or a variety of cloud systems. This strategy requires complete transparency and the ability to gain a birds-eye view into these different systems to provide immediate intelligence as to how the AI came up with its answer. With this ability, organizations can go a long way towards establishing trust in AI results, by verifying where its source data originated.

Strategy 2: Prioritize Better Data over More Applications: Being able to verify AI results is critical, yet what if AI results are so poor that they do not justify analysis? Who takes the blame at this point? It could be that there is something amiss in the programming, but often the problem lies in the quality of the data. If an AI lacks the data to answer a given prompt, or if it is given incorrect data, it will “hallucinate,” or offer an inaccurate response.

One of the persistent challenges around maintaining data quality lies in the fact that very often, each individual application has its own siloed data source. For this reason, the more applications proliferate, the more data can “hide in a corner” and never make it into an AI analysis, or worse, they can create conflicting versions of different data sets.

This strategy is to prioritize the quality and accessibility of the data over simply accumulating more applications. Ideally, the application layer can be decoupled from the data layer, so that the data, independent of any application, can be accessible for analysis and evaluated for its quality and completeness across the enterprise.

Strategy 3: Make AI Development Iterative: Technology moves quickly, but this is especially true in the arena of AI development. New large language models (LLMs) emerge frequently, and new AI capabilities and approaches are announced on a regular basis. In this type of climate, an iterative development process provides your best chance of success, but what would this entail?

It requires most data to be available at the moment it is needed, whether that means next day, next hour, or in real time, while also preserving data security and data governance. It means being able to quickly apply the latest AI models and creating feedback loops between the business and technical teams. With this capability, organizations can immediately address any issues that come up in development, rapidly test new approaches, and quickly iterate on solutions to see tangible results.

The Right Data Foundation

To establish a data foundation that can enable all three of these strategies, a flexible, powerful solution is a logical data management platform. Such a platform can connect disparate data sources to enable instant views of the complete lineage of any data and provide real-time access to data across different sources, all of which critically supports iterative AI development. Logical data management platforms can work on their own or alongside existing data lakehouse or cloud data warehousing solutions, leveraging previous or existing investments.

Key Examples of AI in Energy Applications: By bringing AI into the energy equation, supported by the right data foundation, companies can enable a wide range of solutions, including:

• Power grids that can automatically trigger predictive maintenance and quick responses to outages.
• More effective prediction of solar and wind yields to dynamically balance supply and demand.
• Systems that can immediately leverage diverse customer data for personalized recommendations to improve efficiency and customer engagement.
• Automated environment, social, and governance (ESG) tracking, guided by AI, to facilitate ongoing, dynamic compliance.

The Way Forward

AI can be a boon for energy companies, empowering them to navigate complex challenges and seize new opportunities. However, this potential can only be fully realized if they can provide a robust data foundation that enables transparency, data quality, and iterative development essential for effective AI. Logical data management platforms provide a flexible way to establish such a foundation, supporting existing systems and ultimately contributing to a more resilient, efficient, and profitable energy supply chain that directly benefits power generation.

—Quinn Lewis is vice president, Customer Success Services at Denodo, a data management group.