Ethical AI Use in the Energy Sector: Insights from Infosys EVP Ashiss Kumar Dash
The sun doesn't always show up and the wind doesn't keep a schedule. That variability is the core issue holding back grid-scale renewable integration. Ashiss Kumar Dash, EVP & Global Head Services, Utilities, Resources, Energy & Enterprise Sustainability at Infosys, lays out a simple path: apply AI to forecast better, plan smarter, and operate with discipline-while putting responsibility first.
Infosys has delivered more than 60 AI projects in energy. One deployment for a major North American utility pushed load forecasting accuracy to 95% and avoided tens of millions of dollars in unnecessary infrastructure spend. The message to executives is clear: AI isn't an experiment anymore; it's a lever for reliability, cost, and compliance-if you build it on the right foundations.
Where AI Actually Moves the Needle
- Weather and load forecasting: Better short- and long-term predictions improve dispatch, storage, and capital planning. Research and practice continue to advance in areas like solar forecasting.
- Predictive maintenance: Detect failure signatures early and cut downtime on generation, substations, and network assets.
- Grid modeling and design: AI helps evaluate scenarios quickly, so planners can prioritize upgrades where they matter most.
- GenAI for operations: Field safety checks, automated reporting, customer virtual agents, and engineering support via digital twins.
- Decision support: From supply chain to finance, AI surfaces insights that keep performance and sustainability in balance.
Make AI the Strategy, Not a Side Project
Dash's point: there shouldn't be a separate AI strategy-the AI strategy is the business strategy. Build a foundation that scales across the enterprise and holds up to regulatory scrutiny.
- Tech incubation: Test new tools against real energy problems, not lab demos.
- Platform and ecosystem: Standardize data, architecture, and integrations so models don't get stuck in silos.
- Responsible AI: Set guardrails for fairness, transparency, and compliance from day one.
- Right use cases first: Prioritize high-ROI problems like downtime reduction, planning accuracy, and automation payback.
- Reusable solutions: Create templates and models that work across engineering, finance, HR, and legal.
Execution matters more than slideware. Infosys forms "strategy and execution pods" that combine tech, domain, and business expertise to build, test, and scale fast. The framework stays flexible as regulations and tech shift, but the standard of measurable impact does not.
Case Study: Utility-Scale Forecasting at 95% Accuracy
A major North American utility used Infosys AI models for on-demand, bottom-up, long-term grid forecasting. The stack combined open-source ML, a CIM utility model, and scalable cloud architecture.
- Load forecasting accuracy improved to 95%.
- Saved tens of millions (USD) by moving away from "annual peak load" planning and cutting unplanned outages from >5-10% to <1%.
- "Forecast at will" for multiple horizons: 10 years (DSP), 15 years (scenario study), 22 years (2045 study).
- Processing in under 24 hours for 650K transformers vs. 85 hours for 6,500 assets-25x faster while using 100x more nodes.
- Greater agility in responding to emerging regulatory needs.
The takeaway: when data, models, and architecture are aligned, planners can redeploy capital with precision and confidence.
Responsible AI Is Non-Negotiable
AI earns trust when it's fair, explainable, and compliant. That requires more than good intentions; it needs governance that starts before the first model goes live and continues through monitoring and audit.
- Governance and risk: Establish clear ownership, model documentation, approval workflows, and audit trails mapped to frameworks like the NIST AI Risk Management Framework.
- Bias and fairness: Define fairness metrics per use case, test regularly, and retrain with diverse data.
- Transparency: Right-size explainability for regulators, customers, and employees-especially for decisions that affect billing, outages, or safety.
- Continuous monitoring: Watch for drift and unintended outcomes; tie alerts to incident response.
- Cross-functional oversight: Bring in legal, HR, compliance, and operations early, not after deployment.
- Enablement: Train teams so they know what AI can do, where it fails, and how to use it responsibly.
Practical First Moves for Executives
- Appoint a small "strategy and execution pod" with authority to ship production use cases.
- Pick 3-5 high-impact problems: forecasting accuracy, outage reduction, asset life, and customer operations.
- Stand up a common data layer and model registry so teams don't reinvent the same solution five different ways.
- Create a simple, enforceable responsible-AI policy with model cards, review gates, and monitoring standards.
- Prove value fast with a forecasting or predictive maintenance pilot; scale once you hit targets.
- Upskill leaders and practitioners so adoption sticks and risk stays managed. If you need structured paths by role, see AI courses by job.
The Bottom Line
AI can make renewable-heavy grids more reliable and operations more efficient. Dash's perspective is pragmatic: treat AI as core strategy, build reusable systems, and hold the line on responsibility. That's how you turn intermittent resources into dependable outcomes-and keep regulators, customers, and boards aligned.
Your membership also unlocks:
