A conversational AI agent on WhatsApp for an SMB in Colombia costs between COP $8M and $25M for a 4-6 week MVP, depending on intent count, systems to integrate and personalization level. Average monthly support is COP $800K to $2.5M. This guide breaks down each cost component so you can budget without surprises.
Prices vary dramatically by provider: a local agency (Neiva, Bogotá) charges less than a multinational, and a freelancer even less — but with higher technical risk. Below, actual 2026 market ranges.
Price ranges by agent type
| Agent type | MVP | Timing | Monthly maintenance |
|---|---|---|---|
| FAQ bot (fixed answers, no integrations) | COP $3–6M | 2-3 weeks | $300–600K |
| Sales/support agent (CRM + DB) | COP $8–15M | 4-6 weeks | $800K–$1.8M |
| Multi-flow agent (orders, scheduling, payments) | COP $15–25M | 6-10 weeks | $1.5–2.5M |
| Voice + WhatsApp + web + mobile | COP $25–60M | 10-16 weeks | $2.5–5M |
What's included in each range
Development (one-time)
- Conversation design (prompts, intents, fallbacks).
- WhatsApp Business API integration (Twilio, Meta Cloud API, 360dialog).
- CRM/ERP integration (HubSpot, Zoho, Bitrix24, Odoo, Dynamics).
- Vector knowledge base (RAG) if using proprietary docs.
- Operator panel for human handoff.
Monthly recurring costs (NOT from the provider)
- WhatsApp Business API: $50–200 USD/month per volume (Meta charges per initiated conversation).
- LLM (OpenAI, Anthropic, Google): $20–500 USD/month per volume. Typical SMB agent (1-3K msg/day) spends $80-250 USD/month.
- Hosting: $10–50 USD/month (Digital Ocean, AWS Lightsail, Oracle free tier).
Factors that raise the price
- Custom integrations: each legacy system adds COP $2-5M.
- On-premise / sensitive data: private cloud deployment adds COP $3-8M.
- Multi-language: Spanish + English free; adding Portuguese/French +15-25%.
- Compliance (ISO 27001, 42001): documentation and controls add COP $5-15M.
- 24/7 SLA with oncall: doubles monthly fee.
Red flags when quoting
- Fixed "closed" price without discovery — impossible to quote without seeing systems.
- LLM monthly cost not in proposal — passed to client as a surprise later.
- No human handoff — AI agent without escape loses frustrated clients.
- Promises "learns on its own" — no LLM learns without explicit fine-tuning (expensive).
- No monthly maintenance — unmonitored agent degrades in 3-6 weeks.
How to budget correctly
- Free discovery (1-2 meetings): serious providers want to see systems before quoting.
- Deliverable-based proposal: no open hours. Deliverable = concrete functionality.
- Scoped MVP before full scope. 4-6 weeks to validate value.
- Milestone payments (30% start, 40% MVP, 30% production).
Next step
If evaluating an AI agent for your company, schedule a free assessment with Expansión TI. We deliver a priced proposal within 5 business days after discovery. See our service detail at AI Implementation or check out Klerk Agent, our proprietary WhatsApp AI agent product.