How to implement AI in your company
- Pick a process, not a tool. Start with a repetitive task that already has a visible cost: leads waiting for a reply, a question the team answers ten times a day, content that never ships on time. The tool comes later, chosen for that process.
- Measure the starting point. Write down how the process works today: response time, conversion, volume, hours spent. Without that number, there's no way to prove the AI helped.
- Write the rules and limits. Define what the AI does on its own, what it never does and at what point it hands off to a person. Discounts, promises outside policy and serious complaints usually stay with the team.
- Prepare the data. Gather the information the AI will use — policies, prices, FAQs, CRM history — and name who keeps it all up to date.
- Start small and review. Release one use case, read a sample of what the AI produced every week, correct it and only then expand.
- Train the team and write a usage policy. People who work alongside the AI need to know what it does and how to correct it. A short policy states which data can go into AI tools and which can't.
- Measure in revenue and decide the next step. Compare against the starting point. If the metric improved, expand; if not, adjust the process before switching tools.
Where AI fits in a company
Nexo works in the three areas where AI moves revenue fastest. They can be deployed separately, but they deliver more when they share the same CRM and the same data.
| Area | What the AI does | How it's measured |
|---|---|---|
| Sales | Qualifies leads on WhatsApp, books meetings, follows up and analyzes sales reps' conversations | Response time, conversion rate and sales per rep |
| Customer service | Answers frequent questions, triages requests and hands off to a person with the history | Confirmed resolution, requests for human support and satisfaction |
| Marketing | Produces content, structures the website to be cited by AI engines, generates creative variations and optimizes media spend for sales | Cost per customer and revenue by channel and campaign |
Why so many AI projects never leave the pilot stage
The reasons repeat from company to company, and almost none of them are technical:
- The tool was bought before anyone defined the process it was supposed to run
- Nobody measured the before, so there's no way to show the after
- The AI answers with outdated information because nobody maintains the knowledge base
- The team wasn't trained and goes back to the old way at the first difficulty
- Results are measured in tool usage, not in sales
How Nexo implements
The work follows five stages, the same for all three areas: assessment, process design, deployment, team training and operation with measurement.
Nexo is a sales consultancy that builds its own software. In sales and customer service, the AI runs on CRPRO, Nexo's CRM for WhatsApp, with the conversation, the record and the funnel in the same place. When the company already uses other tools, Nexo integrates them instead of replacing everything.
After the AI goes live, Nexo keeps monitoring: it reads conversation samples, adjusts rules and the knowledge base, and reports on the metric agreed during the assessment.
When it's not yet time to implement AI
AI speeds up the process that exists. If there is no process, it speeds up the mess. It's worth fixing these first:
- When there are no criteria for what a qualified lead is
- When the CRM isn't used and the history lives on each sales rep's phone
- When nobody can say what the current conversion rate is
Usage policy and LGPD
Every implementation includes usage rules: what data the AI can see, how customers are told they're talking to a virtual assistant and how to request human support. Personal data follows the LGPD (Brazil's data protection law) just as it does when people handle customer service.