Caynetic builds
This is the history of an internal Caynetic tool, originally published in March 2026. It is not a client case study. The useful lesson was learning which parts of our communication needed software and which needed a person.
First: a custom AI bot
We built a custom WhatsApp bot to automate replies and route enquiries. As we added branching logic and edge cases, small changes required work in several places. Conversations could feel robotic. A low subscription bill did not remove the maintenance work.
Next: Chatwoot
We moved to Chatwoot for a structured inbox, WhatsApp integration and broader channel support. It gave us a more mature platform. Our everyday need, however, was largely one channel. The question became whether we needed all that scope, not whether open source was good or bad.
Then: a narrower rebuild
We removed the heavy AI reply logic and focused the workflow on WhatsApp with human replies and Slack notifications for inbound messages. The notification helps draw attention to a conversation; it does not replace responding or guarantee that a reply is never missed.
| Stage | Conversation path | Trade-off |
|---|---|---|
| Custom bot | WhatsApp → branching/AI logic → reply | More logic to maintain and review |
| Chatwoot | WhatsApp → structured multi-channel inbox | Mature capabilities beyond our primary channel |
| Narrow rebuild | WhatsApp → human inbox; Slack notification | Smaller scope; human follow-up remains necessary |
What the evidence does and does not show
The original account describes our experience and maintenance trade-offs. This revision adds no response-time, hosting-cost, customer-satisfaction or revenue measurement, because a comparable approved dataset is not available in this audit. It also makes no claim that this March architecture is every component running today.
For a new implementation, test duplicate inbound events, notification failure, assigned-person absence and access revocation. Keep customer content out of broadly visible notifications unless specifically authorised. The right amount of automation depends on volume, privacy and the team’s ability to maintain it.
Try an existing product first, calculate total ownership effort and build only the missing task. The bottleneck worksheet helps make that decision explicit.