JivoChat doesn't build AI in-house, and that's a deliberate design choice rather than a gap. Instead of locking you into one proprietary NLU engine, JivoChat's Bot API lets you plug in a bot provider of your choice — Chatme, Plantt, Aimylogic, or JAICP are the ready-made options — and have it process incoming conversations before an agent ever sees them. For a lot of businesses, one of those four partners is exactly the right fit.
But not every business fits neatly into one of four options, and that's worth understanding before you assume the pre-built partners are your only path.
Why the pre-built list doesn't cover everyone
The four listed bot providers are general-purpose NLU platforms — solid for common qualification and FAQ-style flows, but not built around your specific product catalog, your internal knowledge base, or a particular LLM you've already standardized on internally. Businesses that need any of the following usually find themselves outside what the pre-built partners offer:
- A bot grounded in your own documentation or product data, not a generic knowledge model
- A specific underlying model (say, a particular GPT or Claude deployment your company already uses elsewhere) rather than whatever engine a third-party partner runs
- Deep logic tied to your own systems — checking real-time inventory, pulling an actual order status, or qualifying a lead against your own CRM data rather than generic rules
- Data residency or compliance requirements that a third-party NLU vendor can't meet
None of this means going around JivoChat's architecture. It means building directly against the same Bot API the four partners use — just with your own bot on the other end instead of theirs.
How the connection actually works
The mechanics are refreshingly plain. JivoChat's Bot API operates over webhooks, and JivoChat is always the initiator: when a client sends a message on any connected channel — website widget, WhatsApp, Instagram, whatever's configured — JivoChat POSTs that event as JSON to your bot's endpoint. Your bot processes it however you've built it to (that's the part that's entirely up to you and your chosen AI stack) and sends a response back to JivoChat's endpoint in the expected format. JivoChat displays it to the client exactly as it would a reply from any other provider.
The handoff logic is already built in on JivoChat's side. If a conversation needs a human, an agent accepts the chat and the bot stops receiving events for that thread automatically. Ratings, chat state, and closing events all flow through the same webhook exchange, so a custom bot behaves like a first-class citizen in JivoChat rather than a bolted-on workaround.
What this is, in practical terms
Once you're past the "connect to the API" step, what you're actually building is a real conversational AI application — something with its own model choice, its own grounding in your data, its own logic for when to answer directly versus hand off to a human. It's worth stepping back and seeing how chatbots fit into the broader landscape of modern web development before committing to a build, since a custom bot is really just one piece of a larger conversational strategy.
When it's worth it
If your current setup with one of JivoChat's four partners handles the volume and complexity of what your customers ask, there's no reason to build something custom just because you can. The calculation changes once you notice the bot deflecting fewer conversations than it should because it doesn't actually know your product, or your team spending time working around a generic engine's limitations instead of just using it. At that point, a custom bot built directly against JivoChat's Bot API — with the right AI chatbot development company behind it — usually pays for itself in the conversations it handles correctly the first time.
