Before selling automation to others, we automated our own. We went from manually processing invoices for days to having a flow that classifies, extracts data, and registers everything in the ERP without human intervention. We share what worked, what didn't, and what would have saved us months of iteration.
The unwritten rule in tech consulting is simple: if you don't automate your own operation, don't sell automation. An admin team was spending 80% of its week receiving, classifying, and entering electronic invoices into the ERP. We knew it was the ideal case to test our own tools before proposing them to clients.
Before: someone reviewed the email, downloaded the XML, opened it, copied the fields into a spreadsheet, copied them again into the ERP, and physically filed the document. 20 minutes per invoice. Multiply that by hundreds per month and you understand the problem.
A dedicated mailbox receives the DTEs. A service extracts the XML, validates structure, classifies accounting-wise with rules (if the issuer's RUT is known and the industry is mapped, automatic decision). Only when there's ambiguity does it escalate to an AI model, and only if the AI has low confidence does it move to a human review queue. 92% is processed without intervention.
The most important insight: what we automated wasn't entering invoices, it was the decision of how to classify them. AI added value at the right point in the flow, not as the protagonist but as a tool for a specific case.
Innovating in a 30-year-old company with established processes isn't the same as in a startup. We share what we've seen work (and fail) accompanying Chilean companies in their first AI implementation, and the cultural mistakes that no technical project can overcome.
There's an idea that to implement AI you need a team of PhDs and a huge budget. It's false. Most Chilean SMEs can capture real value with AI without hiring specialists. This is the honest guide on how to start, what to avoid, and how to measure if it's actually working.
After implementing more than 30 automation projects, we identified 5 processes that appear over and over again. We show you which ones they are, how we approached them, and what results you can really expect, with figures from projects in production today in Chile.
Schedule a meeting and discover how artificial intelligence can accelerate your company's growth.
