# From Brief to Live: How We Build a WhatsApp AI Agent in 2 Weeks *Nikhai Jaysen · April 17, 2026* > The step-by-step process behind a two-week WhatsApp AI agent build — what happens at each stage, what we need from you, and what causes delays. Source: https://www.spaceblanket.ai/blog/whatsapp-ai-agent-build-process-two-weeks Most clients have no idea what a two-week WhatsApp AI build actually involves. Here's the exact process — discovery to go-live — and the three things that reliably cause delays. Two weeks sounds tight for a WhatsApp AI agent. Clients are often sceptical when we say it — understandably, since they've read case studies about months-long AI projects, or they've already lived through one. Here's exactly what happens across those two weeks, what we need from you at each stage, and where the whole thing can go sideways when the groundwork isn't in place. ## Stage 1 — Discovery (Days 1–2) In our first session with a new client, we're not selling anything. We're trying to answer one question: what specific job does this WhatsApp agent need to do? That sounds obvious, but it rarely is at the start. Most clients come in with a goal — "handle inbound enquiries," "qualify leads," "help with appointment booking" — but the definition of "qualified" and the step that follows qualification are rarely pinned down. This is what discovery is for. The questions we work through in this session: What does a good lead look like, and what happens to them immediately after qualification? What is the handoff signal — at what point does a human take over? What does your CRM or booking system look like, and who has API access? We also check your WhatsApp Business API status here. If you don't have an approved WABA number, Meta's approval process in India currently takes 3–5 business days. That clock needs to start as early as possible. ## Stage 2 — Architecture and Flow Design (Days 2–4) Once we know what the agent needs to do, we map the conversation logic before writing a line of code. This means a documented flow: every message the bot can send, every branch in the conversation, every fallback, and every integration point. The client reviews and signs off on this document before the build begins. The reason we do this first: a changed conversation branch at day 8 means rebuilding 30% of the agent. The same change as a note in a shared document at day 3 takes 20 minutes to resolve. This stage also locks in the integrations — if the agent needs to push leads into a CRM or pull appointment slots from a booking system, we need API credentials and confirmed access before the build starts, not mid-build. ## Stage 3 — Build and Internal Testing (Days 4–11) The agent is built to spec using the conversation flow as a blueprint. We train it on your specific language — the questions your customers actually ask, the product information they need, the objections they raise — rather than generic prompts. This is where **custom AI development** differs from dropping a chatbot template into a WhatsApp connection: the agent knows your business, not just how to chat. During internal testing, we run roughly 200 test interactions across the expected conversation patterns. Edge cases are deliberately tested: what happens when a user sends a voice note instead of text, what the fallback looks like when a question is out of scope, how a human handoff appears in the WhatsApp thread. We also run the CRM integration end-to-end here — confirming that a completed qualification conversation creates the right record with the right fields populated. ## Stage 4 — Client Review and Go-Live (Days 12–14) You receive access to the staging environment and run your own test conversations. This typically surfaces two things: tone corrections ("it sounds too formal for our customers") and a handful of product-specific questions we hadn't trained for. Both are fast to resolve — usually within a day. On go-live day, we flip the agent from staging to your live WABA number, monitor the first 48 hours of real conversations, and hand over the management documentation: how to update product information, how to read the conversation logs, and how to flag anything that looks wrong. The **workflow automation services** don't stop at deployment — the first two weeks post-launch are the most important for catching what the test suite didn't. ## What Actually Goes Wrong Three failure modes that reliably cause delays: - **WhatsApp Business API approval isn't in place.** We need a live, approved WABA number to connect the agent. If you come to discovery without one, add 5–7 business days to the timeline. We can help you apply, but the Meta approval clock starts from day one of the application. - **Scope expands mid-build.** "Could it also handle post-purchase support?" is a legitimate question — but at day 8, it restarts part of the flow design. We hold a check-in at day 5 to catch additions early, but anything after that pushes go-live. - **CRM access takes time to grant.** In one engagement with a real estate developer, the CRM admin was on leave for a week mid-build. The agent was ready at day 9; the integration didn't go live until day 16. Granting API access before the build starts is the single easiest way to protect the timeline. We've built exactly this — scoped, integrated, live in two weeks — for real estate developers, healthcare clinics, e-commerce brands, and B2B services businesses across India. [Get in touch](/contact) and we'll show you what it looks like for yours. ## Related - [All articles](https://www.spaceblanket.ai/blog) - [Contact](https://www.spaceblanket.ai/contact)