Callbook AI Taught a Machine to Make the Call Nobody Wants to Make
Two founders from Colombia looked at the drudgery of chasing overdue invoices and built voice agents to do it - in Spanish, English, and Portuguese. This summer, Y Combinator agreed the problem was bigger than one country.
The collections call is one of the small, universal miseries of running a business. Someone owes you money. Someone has to pick up the phone. The conversation is awkward on both ends, it rarely goes to plan, and a call center full of people burns out doing it all day. Diego Avellaneda and Daniel Martinez, two founders from Colombia, looked at that exact chore and decided a machine should make the call instead.
The company they built, Callbook AI, makes AI voice agents that dial customers and hold a real conversation - reminding them a payment is due, confirming an appointment, qualifying a lead, or answering a basic question. It speaks Spanish, English, and Portuguese. And this summer it was accepted into Y Combinator's S26 batch, the same accelerator that seeded Airbnb, Stripe, and Dropbox.
01 / WHAT IT DOESA phone system that dials itself
Callbook AI started as something narrow and useful: an AI phone system that plugs directly into Zoho CRM. Instead of hiring developers to stitch together a dialer, a voice model, and a database, a company connects Callbook and lets it work off the CRM it already uses. A workflow in Zoho triggers a call. The agent talks to the customer. Then the system transcribes the conversation, pulls out what matters, and writes it back into the right CRM fields - no human typing up notes afterward.
The agents are meant to sound like people, not menus. Accent, tone, and the way a call is handled can be tuned to fit a business, and the system knows when to hand a call off to a human. What began as a Zoho add-on has since grown into a multi-channel platform that reaches customers by phone, WhatsApp, SMS, and email from one place.
02 / THE FOCUSWhy collections, of all things
Plenty of startups build "AI that calls." Callbook's sharpest bet is where it points that technology: debt collections and portfolio management, first across Latin America. It is not a glamorous corner of software, which is exactly the point. Traditional collections runs on manual scripts and call centers with low contactability - lots of dials, few connections, and staff who dread the work.
Callbook works the whole timeline rather than only the angry late-stage call. It handles preventive reminders before a due date, captures payment commitments in real time, and follows obligations through recovery up to about 90 days past due. The agent adjusts its approach to the debtor's profile, which is where the balancing act lives: push too hard and you lose the customer, too soft and you lose the money.
The collections timeline Callbook works
03 / WHO USES ITSales teams, lenders, and the finance office
The customers are small, medium, and large businesses that live and die by the phone - lenders and finance teams needing to collect, and sales teams needing to reach leads. Beyond collections, the same agents do lead prospecting, appointment confirmation, and customer-loyalty calls. Early client references include the Colombian consumer-credit firm Sistecredito. The team is small, roughly ten people, and it builds its own voice technology rather than reselling someone else's model.
04 / THE MODELYou pay when the phone is answered
Callbook's pricing is unusually honest for software. There is a monthly platform fee - starting around $100 - and then a charge per connected minute of call time. If nobody picks up, there is no per-minute charge; billing kicks in only when a call is answered or reaches voicemail. Higher tiers add AI insights, sentiment analysis, and enterprise customization. The company has been largely self-funded and, ahead of YC, was preparing to raise for product and commercial expansion.
05 / THE EDGEOwning the voice, and knowing the market
Two things separate Callbook from the crowded field of voice-AI tools like Retell, Convin, Tovie, and Floatbot. The first is ownership: the founders say they built their own voice technology, which lets them bend accent, tone, and behavior to each client instead of accepting whatever a third-party model offers. The second is depth in a market most of Silicon Valley overlooks. Callbook did not launch in English and hope Latin America would follow - it started in Spanish, where the collections pain is real, and let global relevance find it. Its real competition is often not another app but the traditional call center itself.
There is a strategic tidiness to the story worth noticing. Callbook did not open with a grand vision. It opened with a wedge - a Zoho CRM add-on one customer would pay for - and expanded from there into a platform. The founders have said the longer arc runs past business calls entirely, toward health, education, and digital-inclusion projects, though for now the focus is squarely on getting paid.
06 / WHERE IT FITSThe quiet corner of the AI boom
Most of the attention in AI goes to chatbots and copilots. Callbook AI sits in a quieter corner: the phone. It is a reminder that some of the most valuable AI companies may not be building anything you type into - they may be building the thing that answers when it rings, in a language a call center struggled to staff, for a job nobody enjoyed doing anyway. Y Combinator's bet is that the corner is a lot larger than it looks.