Your team already uses ChatGPT, each person in their own way. Here is how to go from scattered use to a first AI workflow that saves hours you can measure.
How many Colombian companies already use AI?
According to Cintel, about 40% of Colombian companies already use some AI tool. Using it is not the same as getting value from it: only around one in ten has AI built into business processes, and another 40% say they have no plans to adopt it.
In practice it looks like this: the manager uses ChatGPT for emails, someone in sales uses it for quotes, and nobody knows which customer data is being pasted in. There is usage, but no method.
Which process should you automate first?
Your first project should not be the most ambitious one. It should be the most boring. Look for a task that meets three conditions:
- It repeats every week, ideally every day.
- It follows clear rules someone could write on one page.
- It can be measured: hours spent, errors, response time.
Typical examples in a small business: answering the same WhatsApp questions, moving data from invoices or PDFs into a spreadsheet, drafting contracts from a template, or summarizing meetings and assigning the follow-ups.
How do you know AI is actually saving time?
Measure before you start. For one week, log how long the task takes and how many errors or redos it causes. With that baseline, any improvement can be checked instead of assumed.
At 77Rentals we did exactly this with lease contracts: we timed how long each one took by hand and compared it with the new flow, which generates the document and sends it for e-signature. The difference shows up in the calendar, not in a slide deck.
What does Colombian law say about AI and customer data?
In Colombia, personal data is governed by Law 1581 of 2012. On top of that, the Superintendence of Industry and Commerce (SIC) issued External Circular 002 of 2024 with guidelines for handling personal data in AI systems.
In day-to-day terms: decide which information may go into an AI tool and which may not, use business accounts instead of personal ones, and write down what you use the data for. That is not red tape; it is what protects you when a customer asks.
Do you need to hire a developer?
Not always. Many workflows can be built with tools you already pay for, like Google Workspace or Microsoft 365, plus a well-configured AI assistant. When code is needed, tools like Claude Code or Cursor let a small team build in weeks what used to take months.
What you do need is someone who understands both the process and the technology. That is usually the missing piece.
A 30-day plan to get started
- Week 1: list the repetitive tasks and measure the three that take the most time.
- Week 2: pick one, write down the process and define which data AI may touch.
- Week 3: run a pilot with two or three people on the team.
- Week 4: compare against the baseline and decide whether to scale or adjust.