Use Case Intelligence RSS Search use cases Search
IBM Energy Brazil

CEMIG

Customer Service Automation

CEMIG needed to improve service quality and accessibility for millions of customers across digital and traditional channels, especially in rural and underserved areas where preferences, infrastructure, and digital literacy vary widely.

Read the original story Explore all use cases
1.1 million monthly WhatsApp messages
20% reduction in AHT of branch agents
20-point increase in Net Promoter Score
30% decrease in OPEX

Reported impact

Business outcome

CEMIG raised digital engagement and customer satisfaction, lowered operating expense, reduced claims-related compensation costs, and created a more scalable service model across both digital and physical channels.

  • Compliance
  • Cost Reduction
  • Customer Experience
  • Scalability

Solution

CEMIG partnered with IBM Consulting to redesign customer engagement around AI-assisted, multichannel service. IBM used watsonx.ai as the core intelligence layer for virtual assistants, then combined it with open-source large language models such as Llama and Hugging Face plus NeuralSeek enhancements to improve response accuracy and contextual understanding. The team also unified the utility’s mobile app, online portal, WhatsApp channel, and related touchpoints so customers could move through a more consistent digital journey rather than encountering separate channel experiences. Those changes were paired with process optimization and self-service expansion, which let the utility serve a large, diverse population while still maintaining physical-branch service for segments that need it. The result is a customer-engagement model that uses agentic AI to blend digital assistance, channel integration, and back-end efficiency—improving satisfaction, reducing operating cost, and making digital support more inclusive across different customer groups.

Data flow

Customer interactions from WhatsApp, the mobile app, the online portal, and other service channels feed the watsonx.ai-based assistant layer, which combines open-source LLMs and NeuralSeek context handling to generate responses and then routes customers toward self-service completion or branch/agent follow-up when needed.

Solution architecture

6 components · 3 layers
  1. Compute
    • IBM watsonx.ai Provides the intelligence layer used to develop and improve CEMIG’s virtual assistants.
    • Llama Serves as one of the open-source large language models supporting the assistant stack.
    • Hugging Face Provides open-source model resources used alongside watsonx.ai in the virtual-assistant solution.
  2. Serving
    • IBM watsonx Assistant Delivers customer-facing virtual-assistant experiences across CEMIG’s digital service channels.
  3. Orchestration
    • IBM Consulting Designed and executed the broader digital transformation and customer-engagement strategy for CEMIG.
    • NeuralSeek Enhances the assistants with more accurate and context-aware generative responses.

Architecture clues

  • Digital channel integration spans WhatsApp, mobile, web, IVR, and physical-branch interactions.
  • The assistant stack combines watsonx.ai with open-source LLMs and NeuralSeek.
  • The transformation deliberately kept traditional service channels active for specific audience segments.

Evidence from the source

“30% decrease in operational expenses (OPEX)”
“A 20-point increase in Net Promoter Score”
“Central to this transformation was the use of IBM watsonx.ai, which helped with the development of intelligent virtual assistants and the integration of open-source large language models (LLMs) such as Llama and Hugging Face.”
“IBM also assisted CEMIG in merging multiple digital platforms, such as their mobile application, online portal and Meta-owned WhatsApp”
“These AI capabilities were further enhanced by NeuralSeek”
Transparency Evidence & methodology 100/100 reference value · 89% confidence · 1 source

Review how this reference was scored and trace each normalized claim back to its documented source.

01

Scoring methodology

How scores are calculated

Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.

Customer reference value100/100
  • Problem clarity20/20
    Measures whether the customer's operational problem is specific enough to compare.
  • Implementation depth25/25
    Rewards named components, data flow, integrations, and architecture detail.
  • Outcome specificity20/20
    Measures whether the reported benefit is concrete rather than a generic claim.
  • Quantified impact25/25
    7 metrics describe measurable impact.
  • Deployment scale10/10
    Rewards evidence of a live or scaled deployment and meaningful integration scope.
Evidence confidence89%
  • Source authority20/25
    The main evidence is an official vendor-published customer story.
  • Claim traceability21/25
    10 of 12 key fields link to a named source.
  • Direct evidence20/20
    5 direct source quotes retained.
  • Date provenance10/10
    The publication date comes from the source page or publisher.
  • Source transparency8/10
    1 source reference document publisher, URL, authority, retrieval, and supported fields.
  • Implementation specificity10/10
    Rewards source-backed components, data flow, quantified impact, and a specific outcome.
02

Traceability

Sources & provenance

Field coverage shows which facts each source supports. Dates are shown only when documented by the source.

  1. IBM

    CEMIG customer story

    First-party
    https://www.ibm.com/case-studies/cemig
    Source type
    Vendor-published customer story
    Published
    Dec 3, 2025
    Supports
    • Customer
    • Business problem
    • Solution
    • Products
    • Solution components
    • Data flow
    • Integrations
    • technical area
    • use case category
    • Business outcome
    • outcome category
    • Metrics
    • Architecture
    • evidence quotes
    • Publication date

Date evidence

Not separately sourced Industry, Region, Company size. Treat these normalized attributes as lower-confidence until a dedicated source is added.

Related

Similar use cases

See all like this
IBM Jan 10, 2025 AGL Energy Energy Australia AGL needed to offer always-on digital support across web, account and mobile channels while reducing the load on frontline agents and enabling more transactions to be completed in self-service. 125,000 customers each month are starting a chat Customer Service Automation IBM watsonx Assistant IBM May 14, 2024 E.ON UK Solutions Energy United Kingdom E.ON needed a scalable and flexible replacement for a legacy contact center so it could support growing green-energy demand without disrupting customer service. 1,000 agents Customer Service Automation AWS Cloud · Amazon Connect · IBM Consulting Services for AWS Cloud IBM May 14, 2024 Towngas Energy Hong Kong Towngas needed to handle surging customer-support demand, language complexity and longer call waits while reducing the labor intensity of its contact center during the pandemic. 100% reduction in customer wait times Customer Service Automation IBM watsonx Assistant · SYVA AI chatbot platform