Megacable is one of Mexico’s largest telecommunications providers, supporting customers nationwide. To ease the workload on its call center, Megacable’s independent AI department built AI voice agents for support, sales and collections calls. However, business teams previously waited weeks for IT tickets to resolve data issues, while legacy APIs took 1 to 2 minutes to provide customer context to live AI agents. By adopting the Databricks Data + AI Platform on AWS, Megacable uses Genie Agents and Genie One to help contact center teams find what drives calls, finance teams spot revenue anomalies and sales teams get consistent KPIs. Megacable also uses Lakebase Postgres to deliver subscriber context to AI agents in milliseconds for under $1,000 a month.
Building an AI-first vision with real-time data access
Megacable’s AI department operates independently from traditional IT, driving an organizational mission to empower corporate staff and field teams with AI. The primary focus is Megacable’s nationwide call center, where hundreds of representatives handle support, sales and collections calls. To modernize operations and lighten the load on human representatives, the team set out to deploy autonomous AI voice agents capable of conducting natural, real-time conversations with customers. Powering those agents required a low-latency, highly accessible data architecture that legacy systems could not support.
Eliminating metric bottlenecks in IT ticket queues
Before implementing Databricks, business stakeholders across Megacable depended on a centralized BI system managed by IT. Whenever a business leader or analyst needed a new KPI or metric, they had to open a support ticket and wait while IT prioritized the request against competing backlogs. This bottleneck kept data ownership locked within IT and prevented business stakeholders from exploring the data they understood best.
"Every stakeholder had to ask several people for a metric, and the answer could take days, weeks or even months. We wanted the business to own its data and reach it the right way," said Saul Romo Orozco, Data & AI Governance Lead at Megacable.
Overcoming API latency during live customer interactions
The AI voice agents faced a separate challenge in the call center. The legacy IT APIs that supplied subscriber data such as payment history, active promotions or contract details took 1 to 2 minutes to respond. During a live voice call, that delay leaves customers waiting, and many hang up before the agent can answer. The legacy APIs also lacked autoscaling, so they stopped responding when call volume spiked and left agents without the context they needed.
"If a customer asks the agent something and the data arrives a minute later, they hang up. In a voice conversation, there's no second chance," said Saul.
Empowering business teams with Genie Agents and delivering real-time context with Lakebase Postgres
To address both self-service analytics and real-time agent performance, Megacable deployed the Databricks Platform on AWS as its foundation for governed data and real-time AI. Megacable became one of the first companies in Mexico to adopt Lakebase Postgres. Data moves from source systems through bronze, silver and gold layers before being published to Lakebase. This architecture provides low-latency access for AI agents while allowing business stakeholders to self-serve data through natural language.
Contact center and finance teams find call drivers and revenue risk with Genie
Rather than building endless static dashboards, Megacable’s data team curates high-quality tables in Unity Catalog and trains business data owners to work with their data using Genie Agents and Genie One. Non-technical teams create targeted Genie Agents for specific operational domains, while Genie One serves as a shared source of truth across the business.
In Genie One, the data team builds skills that pair business context with the correct data sources, so common questions get a standardized answer. The sales team, for example, uses skills to pull routine customer KPIs such as customers per region without searching through dashboards. Whether the CEO or any other business stakeholder asks, Genie One returns the same answer.
“True data quality requires delivering consistent outputs for the same question, regardless of who is asking,” said Saul.
Contact center teams use a Genie Agent to analyze call data, which revealed that requests to update account details drive the highest call volume. Finance teams use Genie Agents alongside revenue assurance dashboards to spot revenue anomalies. Data engineers use Genie Code to speed up development, drawing on its built-in understanding of Megacable’s notebooks and platform objects. Beyond Genie, marketing and B2B sales teams use AI/BI Dashboards that combine national census data with customer records to map prospects by region and exclude existing customers.
"Instead of building every dashboard ourselves, we teach the business to use the tools. That's what lets our team spend less time on requests and more time on governance," said Saul.
Delivering real-time context to voice agents with Lakebase Postgres
To power its live AI agents, Megacable restructures fragmented source data into agent-ready gold tables and publishes them to Lakebase Postgres. AI agents consume this data through low-latency data APIs. A collections agent, for example, retrieves a subscriber’s current balance, debt details and debt relief offers during a call before sending a payment link.
Because Lakebase separates compute from storage and scales automatically, the APIs keep responding during peak calling hours. Megacable exposes only a dedicated API schema in Lakebase and assigns each agent its own service principal, so every agent can access only the API it needs.
"Lakebase scales automatically, so our APIs keep responding even when many customers call at once. Our agents always have the context they need to help," said Saul.
Faster answers for customers and data-driven decisions across the business
Business units across the company now independently explore data, while AI agents deliver reliable, real-time customer support at scale.
Quicker agent responses, lower costs and clearer business insights
The implementation of Lakebase Postgres, Genie Agents and Genie One yielded immediate, measurable improvements across technical performance and business agility:
Millisecond context at low cost: Subscriber context retrieval dropped from 1 to 2 minutes to milliseconds during live calls, while Lakebase serves hundreds of thousands of subscribers and thousands of concurrent users for under $1,000 a month.
Top call driver identified: Contact center teams used a Genie Agent to discover that requests to update account details drive the highest call volume.
Revenue anomalies caught early: Finance teams use Genie Agents alongside revenue assurance dashboards to detect and act on revenue anomalies.
Consistent answers across the company: Skills in Genie One give every stakeholder, from the CEO down, the same answer to the same question, such as customers per region.
Data team refocused on governance: By teaching business teams to answer their own questions and standardizing common answers with Genie One skills, the data team spends less time on requests and more time on governance and deciding which data to integrate next.
Reduced LLM token consumption: Structuring and pre-cleansing context in Lakebase reduced token usage, saving cost during model inference.
"We're not a huge team. Databricks gives us the data foundation for everything we build with AI, and its native tools save us development time we couldn't get otherwise," said Saul.
Expanding Lakebase Postgres and Databricks across future use cases
Building on this success, Megacable is testing Lakebase Postgres as a central store for agent conversation logs, state context and telemetry. Capturing these operations in Lakebase enables change data feed streaming directly back into the lakehouse, eliminating the need for complex external ETL pipelines.
Megacable plans to make Genie One the single source of truth for the entire company. The team is also evaluating Genie Ontology to expand natural language understanding across contact center agents. To further scale its AI ecosystem, Megacable is testing agent self-improvement workflows and evaluating Unity Gateway to centralize model governance, manage token costs and enforce access controls.
"Before Megacable, I hadn't used Databricks. Now I know there was a whole world I wasn't seeing, and we're making it fit Megacable," said Saul.
Megacable on Databricks Lakebase Postgres and Genie
Lakebase Postgres is a managed, serverless Postgres database built natively into the Databricks Platform. It separates compute from storage and scales automatically with traffic. Megacable chose it to deliver structured subscriber context to AI voice agents in milliseconds rather than minutes while keeping monthly costs under $1,000.