Steve Dawson, Founder & CEO at POWERCONNECT.AI.
In my journey of introducing AI to over a hundred utilities and speaking with many more organizations eager to embrace it, I’ve learned that the path to success looks very different from traditional IT rollouts. Instead of a huge, all-at-once deployment, the key is to start small, gather real-world experiences quickly and then build from there.
Why AI Plays By Different Rules
At the beginning, we encountered similar failures—it wasn’t the product itself, but the content. During our test phases with utilities, we discovered how much outdated information lived across websites and internal documentation. It turned out to be a blessing in disguise: It gave us a window of opportunity to clean things up before customer service agents went live or a customer interacted with an AI voice/chatbot.
Most GenAI pilots fail for three big reasons:
Data Readiness: AI is only as good as your data. Outdated content fields bad results.
No Clear Goals: Many jump in without defining success, so the pilot goes nowhere.
Wrong Model Or Setup: Picking the wrong LLM or architecture kills performance.
These lessons highlight that AI success is not about rushing to launch—it’s about thoughtful sequencing, data validation and continuous learning.
Traditional IT projects often involve a “big bang” rollout. AI, however, thrives on quick, iterative learning.
Utilities that begin with their own data and a contained AI assistant—essentially a company-specific ChatGPT for internal services—see immediate value with limited risk.
A Phased Approach That Works
AI success is not about going live everywhere at once. It is about moving forward in stages.
Phase 1: Build Confidence
Start with a simple AI assistant for tasks such as answering internal customer service, IT and field service queries; handling customer FAQs; and assisting employees with daily processes. This helps teams get comfortable while leaders identify which prompts, workflows and data create the most value.
Phase 2: Unlock Real Power
The real transformation begins when AI integrates with billing platforms, payment vendors, outage notifications and other critical systems. This is where real-time visibility and a truly powerful AI-driven experience emerge.
Phase 3: Transform At Scale
Once the foundation is in place and core integrations are running, utilities can:
• Deploy agentic AI that proactively assists agents and customers in real time.
• Launch multilingual, authenticated external chatbots for customers covering outages, payments and service requests.
• Automate complex workflows across departments, from dispatch to customer communications.
• Continuously learn from data to predict customer needs, improve satisfaction and reduce costs.
At this stage, AI shifts from being a supportive tool to a strategic growth driver, enabling utilities to reimagine how they serve customers and operate day to day.
A Foundation That Scales
Once that base is established, the real evolution begins. Agentic AI and advanced external chat solutions can be layered on—multilingual conversational AI that authenticates customers and manages everything from outage updates to payments to self-service queries.
This foundation makes the later phases possible and ensures utilities have a flexible, future-proof AI strategy.
Involving Everyone Safely
For AI to succeed, adoption matters as much as technology. That means involving employees early in testing, running role-based micro-trainings, creating simple guidelines and shared prompt patterns and keeping all data within the company environment for full security.
By the time you reach stabilization, usually after just a few weeks of iteration, you have a solid and ready-to-scale AI solution.
The formula is simple: Start small, focus on real user experience and scale once you have proven the value. By learning and iterating quickly, utilities can avoid the dreaded “pilot graveyard” and deliver lasting impact for both employees and customers.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?