Marco Herrarte – Zubale

Momentum has defined Marco Herrarte’s career long before artificial intelligence became a boardroom mandate. Today, as Information Technology Director for Zubale, he is leading one of the most ambitious AI transformations in Latin American retail logistics. The path to that role began in the disciplined world of systems administration and telecommunications software in Guatemala.

Marco Herrarte | Zubale | Information Technology Director

Herrarte started as a systems administrator for a company building software for telecommunications providers. The work grounded him in infrastructure, reliability, and production realities. In 2013, he joined Bluekite, a startup emerging just as cloud adoption accelerated. Developers were excited by the promise of cloud technology, yet anxious about its complexity and risk. Herrarte reduced that uncertainty through automation and security-first design during the cloud hype era.

Acquisitions followed quickly. Bluekite was acquired by a U.S. company. Xoom was later acquired by PayPal. Within two years, Herrarte experienced two major integrations, each demanding resilience and architectural clarity. At PayPal, he led cloud infrastructure initiatives and managed global scale processes under strict compliance requirements.

“Going through two acquisitions so early in my career taught me that technology decisions must survive change,” Herrarte says.

That lesson anchors his leadership at Zubale.

Building the foundation before the intelligence

Zubale is a retail technology company focused on empowering retailers through on-demand workforce solutions, data-driven logistics, and intelligent operational tools. Operating across six countries in Latin America, the company connects retailers with a flexible workforce while optimizing fulfillment and last-mile delivery through technology.

Upon arriving after the pandemic, Herrarte identified a critical bottleneck: a fragmented cloud environment spanning multiple providers. To unlock the system’s full potential for scale, he and his team successfully designed and implemented a unified, cohesive infrastructure.

Later, he and his team designed and launched Data Zenith, a comprehensive data modernization initiative. Rather than beginning with AI models, he began with the structure.

“There is no smart AI without smart data,” Herrarte says.

Data Zenith centralized ownership, standardized pipelines, and created accessible data streams across the organization. Clean inputs were a prerequisite. By establishing clarity at the data layer, he positioned the company to experiment with intelligence without compromising integrity.

The next step was ZARM, Zubale’s purpose-built AI infrastructure developed in collaboration with EON. Herrarte describes it as a secure AI ecosystem, effectively a business unit in a box. Instead of handing teams raw tools, ZARM provides a curated environment where experimentation, deployment, and scale operate under embedded governance.

Inside ZARM sits a portfolio of approved AI models, including Gemini Enterprise for advanced tasks. It also includes an agentic framework for autonomous agents, an automation engine integrated with tools such as n8n, and secure connectors linking to Data Zenith and external APIs. Every component is configured with guardrails for cost control and compliance. The core idea is to empower employees to experiment with AI while testing emerging tools and trends to maximize productivity.

Governance as the enabler of speed

Scaling AI across multiple countries introduces regulatory complexity. Herrarte rejects the idea that governance slows innovation. In his model, governance enables velocity.

ZARM’s architecture operates on a defense-in-depth strategy across three layers. At the data level, strict residency and sovereignty controls protect regional compliance. A dynamic user-inherited permissions model ensures that any AI agent can access only what the invoking employee is authorized to see. Automated classification systems protect sensitive information.

At the platform level, logic separation isolates country-specific processes so operations in Mexico function independently from those in Brazil. Tools such as Google Cloud Armor provide input sanitization, while immutable audit logs deliver traceability.

Application-level controls complete the structure. Only curated models are available within ZARM. High-stakes workflows incorporate human-in-the-loop checkpoints, guaranteeing oversight where consequences are significant.

“We can move fast because the guardrails are embedded from day one,” Herrarte says.

Financial discipline is equally important. Lessons from earlier cloud cycles taught him that uncontrolled experimentation can generate unexpected costs. Within ZARM, projects operate inside pre-approved budgets with automated prevention mechanisms that eliminate overruns. Strategy precedes technology. Architecture precedes scale.

Transforming operations through predictive intelligence

The practical impact of this infrastructure is visible across Zubale’s technology platform, which powers order fulfillment, workforce orchestration, and digital commerce operations for retailers. AI-powered predictive analytics, fueled by Data Zenith, help optimize order batching, routing, and workforce allocation, improving fulfillment speed and operational productivity for retailers.

“ZARM itself doesn’t contain these algorithms. Instead, it serves as an additional layer enabling us to generate new products and features that will multiply value internally, and for our customers, running on our cloud infrastructure,” Herrarte explains.

Dynamic logistics algorithms analyze historical performance, traffic patterns, and weather conditions to anticipate delays. The system batches orders and recalculates routes in real time, minimizing inefficiencies for retailers and workers alike.

Predictive workforce optimization extends the same intelligence to staffing decisions. Demand signals, promotional calendars, and performance metrics inform forecasting models that match the right worker to the right task based on geography, skill set, and track record.

At scale, the cultural shift is significant. Rather than positioning AI as a replacement mechanism, Herrarte frames it as an augmentation.

“We focus on solving digital friction,” Herrarte says. “When someone gets five hours of their week back, anxiety becomes enthusiasm.”

More than 750,000 automated executions now run each month in ZARM. Many originate from finance, HR, and operations teams. These emerging citizen developers design solutions within ZARM’s boundaries, reinforcing a culture of creation grounded in accountability.

Experimentation with discipline

Herrarte balances experimentation with structured governance through a managed three-stage innovation pipeline. The first stage is the experiment, a low-cost test of a hypothesis. If value appears, the project advances to a proof of concept supported by defined metrics and a business case. Successful proofs graduate to full products. Because every stage lives within ZARM, compliance is native rather than retrofitted.

He credits his background in cloud infrastructure, network, security, and automation as preparation for this chapter. Cloud taught him that high-performance systems require strong foundations. Security instilled a non-negotiable mindset. Automation revealed the limits of linear efficiency and opened the door to intelligent agents capable of reasoning through ambiguity.

Looking ahead, Herrarte hopes Zubale becomes a regional benchmark. He believes the barrier to AI leadership in Latin America is rarely talent. It is implementation discipline.

“AI itself is rarely the problem,” Herrarte says. “Failures usually come from how it is implemented.”

His message to peers is consistent. Modernize data. Embed governance. Then unleash intelligence responsibly. By mastering that sequence, companies can compete globally not only on speed but also on trust.

Outside the office, Herrarte nurtures creativity through reading, music, and hands-on technology projects. Reading sharpens his understanding of complex systems. Music reminds him that structure and creativity are partners. Coding keeps the strategy grounded in practical reality.

Under his leadership, Zubale’s transformation is defined not by hype but by architecture and measurable outcomes. Infrastructure came first. Governance followed. Intelligence emerged as a result. The arc of his career reflects a consistent principle. Sustainable innovation is engineered, not improvised, and that engineering now shapes the future of AI-driven retail across the region.