How long until a WhatsApp agent stops promising what it does not have
Putting a WhatsApp agent that handles reorder and availability, wired to real stock, takes 8 to 14 weeks. The conversation gets good in days. What takes time is the wiring that stops the agent from promising what it does not have.
The part that looks like the project, the assistant that answers well, is the cheapest and fastest. The part that decides whether it serves the operation is the integration behind it.
The realistic range
8 to 14 weeks, narrow scope, a small team. The range assumes a catalog accessible by interface, stock queryable in near real time, and a payment method that already works in the channel.
Without real-time stock, the timeline is not longer; the project is different. The agent becomes a friendly chatbot guessing availability, and the first missed delivery burns the trust it took months to build.
What “data in reasonable shape” means here
This is where the project lives or dies. Reasonable means concrete things.
- Stock queryable by API, updated in minutes, not at end of day.
- A catalog with clean attributes, so the agent recommends the right item.
- A payment path that closes inside the conversation.
- Written limits on what the agent can promise and when it hands off to a human.
The phases
Diagnóstico: 1 to 2 weeks, choose the scope and the criterion. Implementação: 6 to 10 weeks, the agent reading stock, assembling reorder, with the action limit defined. Operação: continuous, the internal team tuning what the agent learns.
The acceptance rule is single: the agent only promises what the system confirms. It is the same as Lu’s.
What moves the timeline by 50%
Three decisions, and you can see all three.
- Scope width. Reorder of known items is fast; the whole catalog is another project.
- Quality of stock access. It is the single biggest factor.
- How many legacy systems join the conversation. Each integration is a negotiation.
The week-one criterion
The number that decides is not “the agent answered well”. It is the rate of tasks completed without a human and the rate of false availability promises, which has to be near zero. An agent that resolves 60% and never lies is worth more than one that tries 100% and errs.
Take the agent project on your roadmap. In ten seconds, name the three indicators that say whether it is working. Stalled? It is a hope, not a criterion. Before building, it is worth seeing where each path wins.
Send me the three candidate indicators your agent will measure. In one hour I will send back which predict customer trust and which predict the broken promise that sinks the channel. If the shape confirms, the two-week Diagnóstico ends with a one-page document, three indicators, three ranges, three triggers, that the squad holds for the rest of the Implementação.