Steps to implement AI in your company.
Six steps, with a real example in each one.
The easiest way to start getting value from artificial intelligence is to bring it into the company's daily work, one process at a time. You pick a process that happens every day. You write down what the team knows about it, and the AI works from that. A person checks everything, and when the AI gets it right, it gets more work.
Bear in mind that this is only the beginning. Implementing AI properly changes how the company runs, and that is a longer project that goes through its strategy. Here we stay with the daily work.
We have done everything below ourselves, at Tec2020: in our own businesses and for a client, an online print shop owned by about twenty printing companies.
1. Pick a process, not a tool
Giving the whole team ChatGPT can help, but each task is still done by the same person. Starting with AI is something else: an AI agent takes over one specific process. An AI agent is a program that reads, answers and gets things done, connected to the company's tools. For example, answering quote requests, picking up the phone when nobody can, or drafting the LinkedIn posts.
The first process should be high volume and low risk: something that happens every day, where a mistake can be fixed with another email. Why we always start this way is in the article «AI virtual employees: what they are and when they pay off».
Our AI receptionist platform, Recepción24, started with the phone and the WhatsApp of Reparaman, our home repair service in Madrid. At the print shop, the first process was answering quote requests.
2. Look at how the job is done today
Before you change anything, find out where the work comes in, in what shape, and how much of it. At the print shop, nine out of ten quote requests arrive by email, each written its own way. So the AI agent starts in the inbox, not in the website form.
Also look at what your current software already does. The print shop's software had payment reminders built in, switched off and in English. We translated them and switched them on, with no AI at all. The full story is in the article «Automating quotes and orders with AI in a small business».
3. Write down what your team knows
The AI only knows what you give it: opening hours, prices, what to do when a customer asks for something that isn't on the price list… If only one person knows it, write it down first. The AI doesn't guess. Without the fact, it either makes something up or says it doesn't know, and the second is what you want from it.
In Recepción24, what each receptionist knows is written in the platform itself, and someone at the business changes it there. Fixing an answer doesn't need a programmer.
4. Have a person check what it does
At first, nothing reaches a customer without a person seeing it. At ListingOK, our company that sets prices for vacation rental apartments, someone approves every sales email before it goes out. At the print shop, the AI agent drafts the replies to customers, and someone on the team reads them, fixes them if needed and sends them.
On the phone or in a chat there is no time to check before the answer. There the check comes afterwards: every conversation is saved so someone can read it. And when the AI is unsure, it stops and hands the conversation to a person.
5. Measure and correct
Once the process runs, look at the numbers: how many requests the AI resolves on its own, how many it hands to a person, which ones it gets wrong, and what each conversation costs. Often the fault goes back to step 3: a missing fact or a badly written rule. Fix it and look again. The AI needs supervision, and that doesn't end on day one.
6. Give it more work when it gets things right
More work can be another process in the same company, another way for customers to reach you (WhatsApp, the website chat), or another business. After Reparaman, Recepción24 took on ListingOK too. And since November 2026 it answers our contact page: Teresa, a Recepción24 receptionist, replies there and can book you a 30-minute call. The platform is the same. From one business to the next, what changes is what the AI knows and what it is allowed to do.
Where we wouldn't start
Where a mistake is expensive or embarrassing and can't be undone. At the print shop, checking the files customers send is repetitive, exactly the kind of work you want to automate. But a mistake there only shows up once the order is printed. So it comes after quotes. The other cases where we wouldn't use AI yet are on the page «What we do».
Where to start
With a list: the jobs your team repeats every day, how often they do each one, and what happens when one goes wrong. Your first process comes out of that list.
If you'd like to go through it with us, write to Teresa on the contact page. We'll tell you honestly what we would automate and what you're better off leaving as it is.
And you'll meet Teresa while you're at it.
Which process in your company would you start with?
The first call is free and takes thirty minutes.
Let’s talk →