Pablo G. Baretto
Pablo G. Baretto here.
Most mid-market companies don’t have an AI problem. They have a process, data, and ownership problem that AI vendors are happy to wrap in a chatbot.
I work with 50-300 person companies that already have at least ERP and CRM, a small IT team and a leadership group under pressure to “do something with AI.” I help them decide what is worth building, what to buy, and what to ignore, so the company gets value from AI.
Typical work takes 4 to 16 weeks:
- I start by understanding the company’s constraints and design an AI stack around them: in-house capacity and skills, privacy requirements, preferred cost structure, existing IT-vendor agreements, and cloud/on-prem footprint.
- Then I implement the first low-risk use cases on the stack, usually an internal chatbot and a RAG knowledge base. The stack is already designed for what comes next: agentic applications, integration with existing systems and anything else the internal team could run themselves. I don’t stay forever and I don’t sell software.
Recent work includes implementing an AI stack for a biotechnology R&D firm. It eliminated shadow AI and reduced research time by around 50% on average with an internal, scientific natural language knowledge base.
Over the last 20+ years I’ve been doing IT solutions architecture, data engineering, and in recent years AI software engineering.