175M+
people message a WhatsApp Business account daily
INVESTOR OVERVIEW
The AI Brain for the business, with a governor on every action.
OnDuty connects what a company already has, its CRM, conversations, orders and documents, into one intelligence layer that remembers, recognizes patterns, and decides what should happen next. Agents are how it acts, not what it is. Every action they propose is evaluated by Selah before it executes, and an action without a clean permit does not run.
THE THESIS
Commerce in LATAM runs on chat. Execution breaks after the conversation.
More than 175 million people message a WhatsApp Business account every day. But revenue and trust leak when conversations fail to turn into payments, bookings, and completions, because the context needed to close them is scattered across systems that do not talk to each other. The AI solutions available today make it worse: according to McKinsey's 2025 State of AI report, 51 percent of companies deploying AI reported negative consequences from inaccuracy. An agent that only sees the conversation in front of it cannot know an account has gone quiet or that the same complaint has come up three times. OnDuty holds the context and governs the action.
51%
of companies deploying AI reported negative consequences from inaccuracy
17.2M
SMEs in LATAM, 99.5% of all businesses
MARKET OPPORTUNITY
$7.75B addressable market across 17.2 million SMEs.
LATAM e-commerce GMV is growing at 9.4% CAGR between 2025 and 2029. Online sales account for only 12 to 15 percent of total retail, compared with 25 to 30 percent in mature markets. Digital payments are fragmented across local rails (Pix, MercadoPago, PayU, ePayco, Bold, Azul), increasing checkout friction and making unified workflows more valuable.
| Country | MSMEs | SAM | Entry phase | ||||
|---|---|---|---|---|---|---|---|
| Colombia | 1.7M | $765M | Beachhead | ||||
| Dom. Republic | ~1.5M | $675M | Beachhead | ||||
| Mexico | 4.1M+ | $1.85B+ | Beachhead | ||||
| Argentina | 1.6M | $720M | Expansion | ||||
| Chile | ~1M+ | $450M | Expansion | ||||
| Peru | ~2M+ | $900M | Expansion | ||||
| |||||||
| Subtotal | Subtotal: ~12M MSMEs, ~$5.4B SAM | ||||||
Beachhead SAM (CO + DR + MX): ~$3.3B
Expansion SAM (+ AR + CL + PE): ~$5.4B
Full LATAM TAM: $7.75B
Chile and Peru are next wave markets, addressable through the same self serve model. 9.4% e-commerce CAGR across LATAM between 2025 and 2029.
THE SOLUTION
Governed execution.
Most AI agents are trusted to act on their own judgement. OnDuty's are not. An agent reads the Brain and proposes an action; it holds no credential and cannot reach anything outside OnDuty. Selah, a separate decision engine, answers permit, hold or deny against the workspace's own policies. Only a clean permit reaches the executor, and a timeout or an error kills the action. A hold goes to a person with the full proposal attached.
An agent proposes. Something else decides.
Agents propose, never execute
Read-only retrieval and one tool that proposes. No credential, no connector, no path to the outside world.
Controlled execution
Every action is idempotent, recorded in a hash-chained audit trail, and decided by Selah before it runs. Today that covers messages, escalations, and the actions a workspace declares on its own endpoint.
Holds go to a person
When a rule holds an action, it waits in a review queue with the proposal, the context and the rule that stopped it. Nothing expires into being sent.
Governance as a separate engine
Selah decides whether a proposed action is allowed; OnDuty executes it. The two are separate systems, so the thing that grants permission is not the thing that wants it, and no agent ever holds a credential.
GO TO MARKET
One platform, staged execution.
Phase 1: Self serve
Fastest path to market.
Webchat, the workspace chat, Telegram and WhatsApp are built; WhatsApp waits on Meta's App Review before customers outside the app can use it. A number connects through Meta's own signup dialog. Target: text-first SMEs across CO, DR, MX, AR.
Growth engine
User acquisition metrics for seed narrative.
Phase 2: Expansion
Deeper workflows, higher ARPU.
Payments wedge live in chat. Sales workflows with lead qualification, handoff, and recovery. Multi-channel coverage. Add-on pricing grows ARPU.
Expansion engine
Increasing ARPU through deeper workflow integration.
Phase 3: Enterprise
Governed autonomous agents.
Guardrails written as policy, a kill switch, shadow mode to prove the rules before they are enforced, and a hash-chained decision trail. Higher ACV, managed deployment.
Product development engine
Case studies, governed workflows, logo credibility.
BUSINESS MODEL
Self serve is the door. Enterprise is the ceiling.
Subscription
Monthly plans from $19 to $99, tiered by seats, connected sources, and model spend, with annual cycles. Extra agents, extra seats and WhatsApp are add-ons on the same subscription. Enterprise is a contract.
Expansion add-ons
Payments in chat, sales workflows, multi-location management, and vertical packs increase ARPU as usage deepens.
Enterprise contracts
Managed deployment, orchestration, custom workflows, and compliance. Higher ACV, longer engagement cycles.
DEFENSIBILITY
Four compounding layers.
01
Outcome-linked data
Payment completion rates, recovery success, resolution rates, conversion, time to completion. Far more defensible than generic chat logs.
02
Evaluation harness
We measure task success and workflow reliability, not just whether the chat sounded good.
03
Vertical packs
Starting with e-commerce recovery and dealership lead conversion, expanding to hotels, restaurants, and more. Repeatable onboarding per vertical and per country.
04
Specialized model component
Model-agnostic today. As we gather consented, outcome-linked data, we incorporate a specialized component tuned for vertical workflows and local context.
TEAM
Built by operators who have shipped in LATAM.
DF
CEO
Daniel Forero
11+ years in go to market and growth across LATAM, the US, and Europe. Prior founder exit in 2018 after raising capital and scaling to 5 countries. Has helped raise over $50M across pre-seed to Series A rounds for clients including operators in mobility, fintech, web3, and commerce. Harvard specialization in emerging markets. 50+ startups advised.
JS
COO
Juan Esteban Sierra
10+ years leading go to market for disruptive technology and corporate innovation projects across Latin America. Partner and Head of Growth at Multiplied. Specializes in GTM execution for emerging tech entering LATAM markets. Owns operations, partnerships, and enterprise distribution across Colombia, Dominican Republic, and Mexico.
FC
CTO
Federico Castanares
7+ years full stack engineering. Production grade AI engineering on Groq and Llama. Builds reliability first execution: validation pipelines, sandboxed environments, and audit-ready architecture.
LET'S TALK
If this thesis resonates, we should talk.
We are open to conversations with investors who understand emerging markets, applied AI with governance, or both. 30 minutes with the founders.