NexaVision AI: what I am building and why it is one of my most important projects

NexaVision AI is one of the projects where I am learning the most because it combines three things I care about: applied technology, product thinking and real business problems. It was born to turn operational chaos into more efficient processes with artificial intelligence, building systems that help small and medium-sized teams reduce repetitive work without losing control over their processes.

The project was founded by Roger Pumarola and me, Gorka Hernandez. The idea is to combine business vision, product thinking and technical execution to create automations that do not stay as nice demos, but can actually run inside a company's day-to-day workflow.

What NexaVision AI is

NexaVision AI is a company focused on AI agents and business automation. We work with flows that connect tools such as n8n, LLMs, APIs, CRMs, Gmail, WhatsApp Business, databases and custom JavaScript logic.

The important distinction is that I do not think of agents as "chatbots". A good agent is not only a conversational interface: it is a piece of software that receives information, understands context, executes a task, leaves traceability and knows when it should escalate to a person.

Inside my portfolio, NexaVision AI also has its own project page: NexaVision AI in my portfolio.

What I have been working on

My part focuses mostly on technical architecture and turning ambiguous needs into concrete operational workflows. Some of the areas I have been working on are:

• Multi-tenant n8n environments, designed to separate clients, credentials, workflows and data.
• Customer support agents, with context-aware answers, brand tone and escalation rules.
• Lead generation and enrichment, connecting public sources, forms, CRMs and scoring systems.
• Content automation, such as blogs, social posts, advertising videos and internal summaries with human supervision.
• API integrations, webhooks and external services so automation does not live in isolation.
• Basic observability, because if a flow fails nobody should have to guess what happened.

One of the main lessons is that the hard part is almost never "calling a model". That is the easy part. The hard part is designing the system around it: clean inputs, quality criteria, fallbacks, permissions, memory, logs and clear limits.

Why n8n fits so well

n8n lets us prototype quickly without giving up a fairly serious architecture. For many clients, the value is not in building a huge platform from scratch, but in connecting the tools they already use and automating the complete flow.

I especially like it because it combines:

1. Visual nodes for common integrations.
2. Custom code when a workflow needs its own logic.
3. Webhooks to connect forms, CRMs and frontends.
4. LLMs to classify, write, summarize or decide next steps.
5. Execution logs to review what happened.

That mix makes development fast, but also explainable. And for a client, understanding what happens inside an automation is almost as important as the automation working.

What I am learning as a product builder

Building NexaVision AI is forcing me to think beyond code. Product decisions matter a lot:

• Do not automate a process before it is properly defined.
• Do not sell AI as magic, but as a concrete tool to reduce friction.
• Create small deliverables that can be validated quickly.
• Measure whether the workflow actually saves time, improves conversion or reduces errors.
• Design maintainable systems, not automations only their creator can understand.

I am also learning that good automation has to respect human work. The goal is not to remove human judgment, but to remove repetitive tasks so people can spend more energy on important decisions.

Where I want to take it

My goal with NexaVision AI is to turn it into a reference project for applied automation in small and medium-sized companies. Not by promising "AI for everything", but by building clear, maintainable and measurable systems.

I want to keep going deeper into:

• Vertical agents by industry.
• Smarter sales and acquisition flows.
• Support systems with human escalation.
• Content automation with editorial control.
• More robust infrastructure to separate clients and deploy faster.

NexaVision AI is, for me, a way to practice software engineering in a real context: talking to users, understanding problems, building, measuring, correcting and iterating again.

You can learn more about the project at nexavisionai.net or reach out through the contact page if you want to talk about AI automation, n8n or agents applied to business.