AI handles chats first
Incoming conversations are answered by AI initially, before any human is involved.
What's shipped, in progress, and planned for the Molecule OS company brain. Each project is an MDX file in content/roadmap — prose lives in git, the site renders it.
Incoming conversations are answered by AI initially, before any human is involved.
The AI hands the conversation to a human when the visitor asks for one.
Working authentication — login and signup in action.
AI answers are grounded in the knowledgebase with inline citations, handled natively through Molecule OS — every reply shows the sources it came from.
A clean, minimal inbox to see and manage conversations, with basic setup — messages from AI and humans update in real time.
Add team members to the workspace.
Add the chat script to any website and the widget loads properly; messages send and receive in real time.
The AI uses earlier messages and known customer details (from HubSpot) to answer in context.
The AI answers when confident, asks to clarify when unsure, and escalates instead of guessing.
Two-way sync with HubSpot: push contacts and conversations into the CRM, and pull customer context back into the inbox.
Add and manage the sources that power AI answers: write articles, upload PDFs, and crawl your website into the knowledgebase.
When the AI can't fully resolve, capture the visitor's details — name, email, and what they need — and push them to the CRM for manual follow-up.
The AI proactively starts conversations based on visitor behaviour and intent (for example buyer vs seller), instead of only waiting to be messaged — to engage more visitors and reduce agent load.
Rerank retrieved content so the AI pulls the most relevant, accurate passages into each answer.
Let the AI fetch data through integrations — order status, subscription, account details — and use it in its answers.
Let the AI book appointments right in the chat, routing to the correct calendar by the visitor's location or team (for example the right franchise office). Supports many calendars and can replace a separate bookings tool.
A single contact view — history, company, and CRM fields — beside every conversation.
Handle support email in the same inbox — AI drafts, humans send.
Run the AI chat inside JustCall so teams can keep their conversations on their existing JustCall workspace.
Pull order and customer data from Shopify into conversations.
Get notified about conversations and reply to them from Slack.
Surface billing and subscription info from Stripe for support.
Two-way sync with Zoho CRM — contacts, conversations, and customer context — for teams standardized on Zoho.
Conversation volume, first-response time, AI-vs-human handling, and trends over time.
Ask customers to rate resolved conversations.
Thumbs up/down on AI replies to catch bad answers and improve over time.
Detect risky or out-of-scope questions and route them straight to a human.
Measure when the AI actually resolved a conversation (confirmed, or no follow-up) — the number that proves the product's value.
Set response-time targets and office hours, and flag breaches.
Stripe-based plans and usage limits so you can charge customers.
Surface questions the AI couldn't answer and suggest new KB articles to fill the gaps.
First-run setup, docs, and reliability polish for public launch.
Extend the chat experience to WhatsApp and SMS.