A buyer at a mining company calls IDICO. He doesn't know which department he needs, so he says the name of the person he's after. Several people at IDICO share that first name, so the voice on the line asks which one he means. A few seconds later he's talking to her.
The voice that sorted that out is an AI receptionist. It has been answering IDICO's front desk since December 2025, in three languages.
IDICO is a strategic sourcing company for heavy industry across Latin America. It replaced a phone menu it didn't control with CloudTalk's AI Receptionist, and its callers now reach the right person or department by asking.
About IDICO
IDICO is a sourcing company for heavy industry in Latin America, going for more than 38 years, with operations across Peru, Colombia, the Dominican Republic and Chile. Its clients are in mining, agroindustry, food, energy, steel, cement and aggregates.
IDICO is a company specialized in strategic sourcing solutions for industry. Our approach combines purchasing, technology and innovation to help clients optimize their processes and generate greater efficiency.
Melissa administers CloudTalk at IDICO: the configuration, the integrations, user support, and working out what else the company can do with it.
Problem
For years, IDICO's front door was an ordinary PBX. Calls connected. That was about all it did.
With the old PBX we had no reporting, no call control, and obviously no AI. We couldn't do anything.
Worse, IDICO couldn't change any of it. Two outside providers held the configuration.
The phone system didn't depend on us. We depended on two providers, we had a lot of questions, and nobody answered them.
Most of IDICO's clients never set foot in the office. The phone is the relationship. And the first thing every one of those relationships hit was a menu.
Evaluation
IDICO's evaluation was run by Melissa's manager, who looked at several tools. One capability decided it.
AI gets a lot of attention at IDICO. We run everything with AI.
What sold them was the idea of a front desk that could simply be asked a question.
The AI receptionist just asks how it can help, and the caller says put me through to this person, or I need a quote. No dialing a number for this department, press this for that. That's basic for a lot of companies, but we wanted something different.
Solution
Building it
The technology was not the hard part. Agreeing what the front desk should say was.
We had to sit down with the owners in Miami, then marketing, quality, processes, IT, all together, to work out how we wanted the project to look and which departments calls should transfer to.
Then came the prompt, which is where most of the real work happened.
The main challenge was really the prompt. Sometimes the agent didn't do what you wanted, but the fix was fast: say something very specific, add a full stop, separate certain things. It was learning how to talk to the agent.
Once the structure was agreed, going live took almost no time at all.
Going live took two or three days at most. What took the time was agreeing the structure.
The moment it clicked
IDICO runs the receptionist in four languages. Spanish, English and Portuguese are live across the business, and Cantonese is configured for the Hong Kong line. The team tested it by trying to break it.
We'd speak to the receptionist in Spanish, then English, then Portuguese in the same call, and it answered. It switched immediately.
They also made it say less. The first version asked callers a series of questions, and IDICO stripped it back to one.
Now the receptionist says: hi, welcome to IDICO, how can I help? It used to ask a lot of things, and we took that out.
What the receptionist does
It finds people by name. A caller asks for someone. If several people share that first name, the receptionist asks which one, then transfers. If nobody picks up, the voicemail arrives in that person's email as audio.
It finds departments. Finance in Miami, finance in LATAM, and so on, without the caller needing to know the structure.
And it hands off to other AI agents. This is the part that makes IDICO's front desk unusual. The receptionist routes to humans, and it also passes callers to three more voice agents sitting behind it, whichever one fits:
- Checking a quoteThe caller gives a quote number and the agent reads back the name, the value, and which insider owns it.
- Requesting a quoteThe agent collects name, email, item and quantity. Power Automate turns the spoken request into a structured email that lands in the right inbox, formatted like a form.
- Raising a complaint or requestThe agent asks whether the caller is a client or a supplier, takes the details, and routes clients to the commercial team and suppliers to logistics and operations.
A caller who wants a price gets it handled end to end, and the request arrives as an email someone can act on.
What runs behind it
The receptionist is the front door. IDICO's Salesforce integration handles what happens after the call connects, so recordings and call activity land in the CRM without a rep typing anything. Power Automate carries the agents' output into IDICO's internal workflows.
Results
Callers get where they're going by asking. They name a person or a department in whichever of three languages they speak, and the receptionist takes it from there.
Quotes stop falling through. Requests that used to depend on catching the right person now arrive as structured emails, whether or not anyone was free to pick up.
The AI agents on our calls mean quotes don't get lost, and the client feels closer to us on a call, because most of them never come here physically.
Forgotten conversations stopped turning into lost deals. A rep who can't remember what a client asked for that morning has the summary and the recording to go back to.
The reps can go back to the summary or the recording, and they stop losing that sale.
Reps stopped doing admin that was never their job.
It comes down to less time on tasks that shouldn't be manual, like logging a call. That gives the reps more time for what matters, which is selling.
Sentiment scoring changed how managers coach. This one came as a surprise to the team.
The happy face and the sad face in the analytics matter to the performance team, along with the call summary saying this call was neutral or this call was positive, because in every one-on-one they can say: I saw these calls went like this.
