FlowbotsAn AI receptionist platform that connects calls, messages, bookings, and follow-ups—giving service businesses a way to keep responding while their people are busy doing the work.
Explore the live productFor a service business, the next customer often arrives while the team is already helping someone else. A missed call can leave a booking unfinished. A new lead needs a response, an existing customer wants to reschedule, and a text may contain information the next person answering the phone needs to know.
Flowbots brings those interactions into one operating system. The challenge was larger than making an agent speak: it needed to understand the business, use the right tools, carry context between channels, and make the resulting work visible to the team.
I built Flowbots end to end: the public product experience and dashboard, business setup, agent configuration, operational APIs, voice and messaging workflows, CRM connections, and the infrastructure that records outcomes and usage. The work joined real-time conversations to the practical details of running a service business.
The home screen brings attention items, calls, leads, bookings, and handoffs into one operating view. Its funnel shows where new leads stop moving forward, while urgent items give the team a place to act. The Messages inbox carries the conversation context into human follow-up: replying takes over from the agent until the team hands it back.
Each agent has a role, instructions, voice configuration, tool access, and a published version. The interface separates saved drafts from the version customers hear and shows where the agent is connected to phone numbers. A visible version history gives operators a way to inspect previous releases and roll back deliberately.
A business owner needs to understand how the receptionist will sound and behave. The voice screen presents presets through practical tradeoffs: naturalness, adherence to the script, response speed, and estimated cost. Advanced controls remain available for a custom setup, while the main comparison makes the decision understandable without needing to know the speech stack.
The first attempt is only part of the workflow. Flowbots supports call-first, text-first, and single-channel approaches, then shows the longer follow-up rhythm as a timeline. Schedule previews make the dates concrete. Replies, bookings, STOP messages, and a teammate taking over end automated follow-up; sending windows and consent rules are handled by the surrounding application.
The quality workspace brings browser testing, simulated customer scenarios, expected behavior, and run results into the agent interface. Tests can cover bookings, handoffs, interruptions, and business boundaries. The platform distinguishes testing a saved draft from changing the published agent, so reviewing behavior is part of the release workflow.
The web application, voice runtime, operational API, and durable event processing connect a live conversation to the business systems that act on it.
A shared dashboard connects setup, agents, calls, conversations, and operational settings. React Query handles server data, while reusable interface components keep the experience consistent across the platform.
Tenant-scoped APIs connect business context, contacts, calendars, conversations, and follow-up settings. The application coordinates work through CRM integrations or its own contact and booking paths.
A dedicated voice runtime manages live sessions, speech, turn detection, interruptions, and tool calls. Published text agents use conversation history and business context to produce replies and carry out supported actions.
Call outcomes travel through a durable outbox, lead outreach advances through persisted jobs, and usage events are recorded with retry handling. Stable event keys help prevent duplicate processing across services.
A live conversation cannot wait for every downstream integration. The voice runtime handles the session while normalized outcomes are delivered through a persisted outbox with retries and a terminal failure state. This gives the operational system a way to recover when another service is temporarily unavailable.
The text runtime loads customer context, recent conversation history, tool results, and incoming photos. Completed jobs return their stored result. If a model fails after a write action, the workflow hands off instead of blindly retrying a booking or saved change.
Consent checks, suppression, sending windows, and lead-local time rules are enforced in application code. Follow-up work can wait for an allowed window, and usage recording uses stable event keys and a durable retry queue. These controls are part of the system around the agent.
Flowbots brings the customer-facing agent and the business-facing workspace together. A business can configure how it responds, connect its operational context, inspect conversations, test agent behavior, and manage the work that follows. The platform spans inbound voice, outbound outreach, text messaging, and human handoff in a common application experience.
The useful part of an AI receptionist is what happens after it understands the caller. A booking needs to reach a calendar, a handoff needs context, and a follow-up needs to respect the customer’s response. Building Flowbots meant making those transitions part of the product.
Captured from the live product and its test workspace. Crops exclude account details, contact records, and wallet balances. Test activity and marketing examples illustrate the interface; they are not reported business results.