Company profile · 3 min read
Callbook AI: multichannel collections agents for late-stage portfolios
AI collections workflow for locating, contacting, segmenting, negotiating with, and supporting borrowers across voice, SMS, and email.
Cofounder of Lazyweb (a16z Speedrun). Previously grew subscriptions and built native ads at Duolingo. Writes First1000 (90,000+ subscribers).
Published · Updated
Callbook AI is an AI collections agency for late-stage portfolios. The buyer decision is whether a lender, fintech, insurer, retailer, or BNPL provider wants multichannel agents that locate, contact, segment, and negotiate with borrowers while preserving a human and compliance control layer.
What it does
Callbook says its agents work across voice, SMS, and email, keep one conversation history, classify ability and willingness to pay, and choose the next action—payment plan, human route, account review, or follow-up. Its developer API can trigger calls, receive webhooks, and retrieve call analytics (Callbook homepage; Callbook developer API).
| Fact | What the public sources say |
|---|---|
| Buyer | Lenders, banks, fintechs, insurers, retailers, and BNPL providers |
| Workflow | Borrower research, segmentation, multichannel contact, negotiation, payment plans, and analytics |
| Public scale claims | Callbook says 23M+ borrowers managed and $11M+ payments recovered |
| Integration | REST API, webhooks, assistant context, and call analytics are listed |
| Founders | Diego Avellaneda and Daniel Martinez |
Why it fits
Collections is not just an outbound dialer. The useful distinction is adapting the next action to the borrower's situation and keeping the conversation continuous across channels. A lender can also use the insights to see why customers fall into debt, rather than treating every account as the same script.
The scale, recovery, and pilot claims are company-reported. A buyer should resolve consent and contact rules by jurisdiction, identity verification, hardship handling, dispute escalation, language coverage, payment-plan authority, recordings, data retention, and how a human can intervene. The product touches sensitive financial conversations; this is an operations profile, not debt or legal advice. Pricing was not published.
Founder fit is technical and product-oriented. YC describes Avellaneda as an applied-ML researcher with embedded-systems publications and automotive e-commerce experience, and Martinez as a former B2B product leader at Treble (YC company profile).
Short version: Callbook is worth evaluating when a lender has a large late-stage portfolio and a clear compliance playbook. Start with a controlled segment and measure recovery, complaints, and escalation quality together.
Sources checked — 2026-09-19
Cohort context
Callbook AI is listed in Summer 2026. In our 2026-09-18 directory snapshot, 119 of 232 listed companies in that cohort have YC’s primary industry label B2B (51.3%). This is a current-directory comparison, not an original intake count or a performance ranking. Nine-cohort dataset.
Public website snapshot
Observed 2026-09-19T16:18:38.472Z in raw homepage HTML. This records visible metadata and advertised links, not agent execution or product quality.
| Signal | Homepage observation |
|---|---|
| Product description metadata | Observed |
| Canonical link | Observed |
| H1 or H2 heading | Observed |
| Typed structured data | Observed |
| Docs/developer link | Observed |
| Pricing link | Not observed in this response |
| llms.txt link | Not observed in this response |
| Markdown alternate | Not observed in this response |
Public observations · Collection method. Missing links here do not establish that a capability or file is absent elsewhere.
About the author
Cofounder of Lazyweb (a16z Speedrun). Previously grew subscriptions and built native ads at Duolingo. Writes First1000 (90,000+ subscribers).
I cofound Lazyweb and publish Mudpie. This is an owner-written publication, not an independent testing organization. Research notes distinguish observations, sourced reporting and editorial judgment.
First1000 ↗ · X ↗