There's no such thing as off-the-shelf AI

There's no such thing as off-the-shelf AI. CV CRM invited me onto its podcast, CV na Prática, and I said what I repeat at every talk: every operation has its own rules, rhythm and vocabulary, and no tool that arrives ready-made in a box knows all three.

That bothers people looking for something to install and forget. But it's the reality for anyone who has tried. A generic agent is right in general and wrong on the detail that matters. It knows how to talk about property, but it doesn't know that at your developer the down payment can be split into twelve installments, that the model unit only opens on Saturdays or that the broker for the south zone doesn't take investor leads.

Getting the tools to talk

The breakthrough isn't one more tool. It's getting the ones you already have to talk to each other: CRM, WhatsApp, customer service. Most sales operations already have all of that. What they lack is real integration between them.

The lead arrives on WhatsApp, the rep jots it in a notebook or a spreadsheet, someone types it into the CRM at the end of the day, the broker gets the record without the conversation, and the marketing campaign never finds out which ad brought in the person who bought. Every system works, but the data doesn't move between them.

Data that doesn't circulate doesn't become decisions. That sentence sums up the problem. There's no point having conversation history if it doesn't reach the broker. There's no point having the customer profile in the CRM if the service agent doesn't check it before answering.

The customer changes along the way

The point that sparked the most discussion on the episode was this: the customer at first contact isn't the same one at signing. Their questions change, their urgency changes, their channel changes.

At first contact, the question is "how much does it cost." Two weeks later, it's "what's the difference between floor plan A and B." Close to signing, it's "how does financing work" and "when can I move in." The urgency changes too: from the browser who saw an ad to the buyer comparing two options who has to decide by the end of the month. And the channel shifts with it: it starts on WhatsApp, goes through the broker's phone call, and returns to WhatsApp to send documents.

Useful AI notices that change and switches the next action, not just the name in the message. Personalizing is easy: drop the customer's name and the project name into the text. The hard part is noticing the person has left the curiosity phase and entered the decision phase, and that the next action is now sending the financing simulation, not another video of the model unit.

The reverse path

Every operation already has its own way of selling, serving and building relationships. The common mistake is asking it to adapt to the software. Our path is the opposite, and it has four steps.

1. Diagnostic to see where the operation gets stuck today, with data, not guesswork. How long the lead waits, where the conversation dies, how many records reach the CRM incomplete.

1. Process design in the company's own rhythm and vocabulary. If the team says "pre-service" and not "qualification," the agent does too. If a visit is booked only with phone confirmation, the flow respects that.

1. Deployment integrated with the CRM and the channels that already exist. Nobody switches systems. The agent plugs into what's there.

1. Ongoing management, because a living operation changes. New launch, new price list, new campaign, new manager. The agent has to keep up.

The full episode is on YouTube and Spotify. Search for CV na Prática.

And if you want to start with step one, the diagnostic is exactly that: looking at where your operation gets stuck, with data.