AI Agent for Energy & Gas Customer Care: A Multi-utility Case Study
An inbound AI agent managed 3,400+ calls, increased the success rate by 13.6 points, and freed up 212.6 operator hours. How can you replicate this in your customer care?
If you work in customer care for an energy and gas/TLC multi-utility, you know the burden of repetitive calls. An AI agent for energy and gas customer care can filter and resolve standard cases, leaving only complex issues for human operators. In this case study, we demonstrate how an inbound AI agent, adopted by an Italian multi-utility, managed over 3,400 calls in just a few months, increasing the success rate by 13.6 points and freeing up 212.6 operational hours. ## TL;DR - A total of **3,402** calls managed (Feb–Jun 2026); **2,177** successfully closed for an overall success rate of **64.0%**. - Growing performance: **57.1% → 70.7%** between April and June (**+13.6 pp**) with a constant average duration (~**3.76 min**). - **3,358** calls in the April–June quarter with a **64.4%** success rate: the most representative period. - **212.6 hours** of automated conversation: reduced operator workload, improved continuity, and standardization. ## Context and Objectives The multi-utility operates in telecommunications, energy, and gas with high inbound volumes: inquiries about contracts, bill clarifications, and service reports. Before AI, many standard requests required human intervention. Project objectives: - Increase inbound call handling capacity. - Reduce customer care operational workload. - Improve service continuity and SLAs. - Standardize responses. - Delegate only complex or sensitive cases to operators. ## Solution: Intelligent Virtual Assistant for Multi-utilities ### Key Capabilities - Understands user intent and follows updated policies and scripts. - Independently resolves simple cases (information, status updates, self-service guidance). - Standardizes responses based on a knowledge base and existing integrations. - Orchestrates conversational AI for TLC customer care with clear dialogue turns and explicit confirmations. ### Escalation and Handoff - Smooth handoff to operators when specialized expertise is needed. - Transfer with full call context to reduce handling times. - Clear fallback rules to maintain service continuity.  ## From Initial Tests to Full Operation Observed period: February 16 – June 30, 2026. Preliminary phase with reduced volumes (8 calls in February, 36 in March). From April, the agent became fully operational. - April: **1,063** calls, **57.1%** success rate. - May: **1,231** calls, **65.4%** success rate. - June: **1,064** calls, **70.7%** success rate. ### Monthly KPIs (Apr–Jun 2026) | Month | Calls | Success rate | |---|---:|---:| | April 2026 | 1,063 | 57.1% | | May 2026 | 1,231 | 65.4% | | June 2026 | 1,064 | 70.7% | | Total Apr–Jun | 3,358 | 64.4% |  ## Measured Results - **3,402** total calls until June 30, 2026. - **2,177** calls successfully completed (**64.0%** overall). - **212.6 hours** of automated conversation (average duration **~3.76 min**). - Success rate growth: **57.1% → 70.7%** between April and June (**+13.6 pp**). ## Why Success Rate Growth Matters More Than Volume Quality remains stable: the average call duration stays between **3.7–3.8 minutes** even with higher volumes. The increase in success doesn't come from longer conversations, but from an optimized flow, better adherence to real-world scenarios, and improved orchestration of conversational AI for TLC customer care. ## Operational Impact and Scale - **>211 hours** absorbed in the April–June quarter: time returned to internal teams. - Standardization of responses and scalability of service. - Efficient prioritization and escalation to human operators. - Tangible value of energy and gas call center automation without replacing people.  ## Conclusions The inbound AI agent has demonstrated a measurable impact on service volume, effectiveness, and continuity, improving month over month. Would you like to test how to apply it to your scope? Request a demo of DeepAgent: you will be called back on +39 068345191. ## Frequently Asked Questions ### What results can an AI agent for energy and gas customer care deliver in 90 days? In the April–June quarter, the system handled **3,358** calls with an average success rate of **64.4%**, growing from **57.1%** to **70.7%** (+13.6 pp). It automated over **211 hours** of conversations. These numbers show how energy and gas call center automation frees up capacity, maintains SLAs, and improves effectiveness without extending calls. ### What types of requests does an intelligent virtual assistant for multi-utilities handle? Standard informational cases: contract details, bill explanations, meter readings, case status updates, appointments, and simple technical reports. The intelligent virtual assistant for multi-utilities follows scripts and policies, verifies data when necessary, and routes complex cases to an operator, ensuring consistent responses and quick turnaround times even during unexpected peaks. ### How does it integrate with CRM, billing, and IVR in TLC customer care? The conversational AI for TLC customer care integrates via API with CRM, billing systems, and IVR for data retrieval, authentication, and ticket updates. The handoff includes the conversation context to reduce Average Handling Time (AHT). The solution adheres to policies and logging, ensuring traceability and standardization of responses at scale. ### How long does it take to go live and see improvements? In the analyzed case, February–March tests had limited volumes; from April, the agent became fully operational, and success rose from **57.1%** to **70.7%** by June. Typically, a pilot implementation requires 4–8 weeks for configuration, training, integrations, and tuning; improvements continue with flow optimization. ### How do I measure the ROI of an energy and gas call center automation project? Define a baseline and monitor: first-contact resolution rate, success rate, deflection, automated hours, AHT, waiting times, service continuity, and CSAT. Also evaluate the qualified time returned to operators for higher-value cases. Connecting these KPIs to operational costs provides a clear and defensible ROI.