WhatsApp AI chatbot with CRM integration
Netbase designed and built an AI chatbot on WhatsApp Business that handles intents and conversation flows and keeps lead and customer data in step with the client's CRM, delivered in milestones from prototype to knowledge transfer. The client, sector, region and CRM product are not named, and no conversation volume or business result is claimed.
Submit a project See the related service
Record at a glance
- Evidence level
- Portfolio record: delivered scope from the project documents, attested by Netbase's CEO
- Client
- A business whose customers use WhatsApp; client, sector and region not named
- Channel
- WhatsApp Business, through the WhatsApp Business API
- Service
- AI Workflow Automation
- Delivery
- Milestones over a four-to-eight-week plan, then 30 days of support
- Published metrics
- None
Context
A business wanted its customers to get answers and start a sales conversation on WhatsApp, the channel they already used, without the details being lost between chat and the sales team. The requirement was a chatbot that could understand what a customer wanted, follow a structured conversation, and write the useful parts, such as a new lead or updated customer data, into the company's CRM.
Netbase role
Netbase designed and built the chatbot and its integrations and handed it over with documentation, knowledge transfer and 30 days of support. The work was delivered in milestones over a four-to-eight-week plan.
Delivered scope
- an AI chatbot on WhatsApp Business, connected through the WhatsApp Business API;
- intent recognition and conversation flows, built with Dialogflow or Amazon Lex;
- language generation through a large language model API (GPT-4 in the recorded stack);
- synchronization with the client's CRM for lead capture and customer data;
- workflow automation triggered from conversations;
- NLP tuning, testing and optimization;
- documentation, knowledge transfer and 30 days of post-launch support.
Technology names describe the project's stack. They do not imply a partnership with any vendor.
How the milestones ran
-
Design and prototype
The conversation flows were designed and prototyped.
-
Model and channel
The language model and WhatsApp were integrated.
-
CRM and automation
CRM integration and workflow automation were built.
-
Tuning and testing
NLP tuning, testing and optimization.
-
Handover
Documentation, knowledge transfer and handover into support.
Separating the channel, the model and the CRM work into distinct milestones meant each integration could be tested on its own before the full conversation was switched on. The project delivery page explains how milestones are defined and accepted.
Evidence level and limitations
- The record describes scope from the project proposals. It does not report conversation volume, conversion, response time or any operating result.
- The documents do not describe a hand-over from the chatbot to a human agent, so this record does not claim one.
- The client, its sector and region, and the CRM product are withheld.
- No screenshots are shown. The image on this page is illustrative and shows no messaging app.
Lessons for a similar project
A messaging chatbot touches customer data and your brand voice, so settle these points before build:
- an approved business messaging account and the channel's rules on templates and opt-in;
- data ownership, consent and retention for conversation data and what is written to the CRM;
- an escalation path: when and how a conversation reaches a person, even if the first release is automated only;
- representative test conversations, including off-topic, angry and ambiguous messages;
- reconciliation when the CRM or the messaging API returns an error or an uncertain result.
Today, AI Workflow Automation projects design the human review point in from the start, and test that the chatbot hands over when it should. The wider support workflow around a bot like this one, from intake to escalation with context, is set out in customer support automation.
Questions buyers ask
Which AI model did the chatbot use? GPT-4 appears in the recorded stack for language generation, with Dialogflow or Amazon Lex for intents. Netbase works with commercial and open-source AI models chosen per project, and naming a model describes the stack rather than a partnership.
Did the bot pass conversations to a person? The project documents do not describe that step, so it is not claimed. A new project should design the escalation path before launch, as the lessons above set out.
How long did delivery take? The plan ran in milestones over four to eight weeks, from design and prototype to documentation and knowledge transfer, followed by 30 days of support.
Planning a chatbot or CRM workflow?
See the other portfolio records or read how evidence is labelled on the methodology page. When you are ready, submit a project brief. OutsourcingVN is operated by Netbase JSC and is Netbase's own outsourcing-services platform.