WHATSAPP AGENT DASHBOARD · CATEGORY · TYPE
WhatsApp Agent Dashboard
WhatsApp Agent Dashboard is a concept product for businesses running customer conversations on WhatsApp. It puts an AI agent directly inside those conversations, then gives a human full visibility into what the agent is doing, with every consequential action, a refund, an account change, an outbound message, surfaced for approval before it happens.
- Engagement
- Self-directed concept project
- Duration
- 1 Week
- Scope
- Product concept · dashboard UX · agentic conversation design
- Year
- 2025
Leads and customer conversations increasingly happen on WhatsApp, not in a CRM. That's where people actually message a business back, a delivery question, a refund request, a "can I change my order." Nothing about that channel is properly tracked.
We noticed this firsthand while doing our own outbound. Leads would come in on WhatsApp and get handled thread by thread, with no system underneath keeping track of what was said, what was promised, or what needed following up.
WhatsApp Agent Dashboard is a concept product built to put an AI agent directly into that channel, not as a chatbot replacing the conversation, but as a dashboard that gives a human full visibility and full control over what the agent is doing there.
This is self-directed work, not a client engagement, the same as FarmOS and Meet Pilot.
This one came out of direct experience running outbound conversations on WhatsApp, rather than a separate round of interviews the way FarmOS and Meet Pilot had.
Grounded in personal experience running outbound, not a formal research study.
The conversation is the record
Nothing about a WhatsApp thread gets pulled into a system automatically. Whoever's handling it has to remember what happened, or lose it.
An agent needs a way to be wrong safely
Refunds, account changes and outbound messages are all things a business needs to approve, not something an agent should just do on its own.
Confidence isn't the same everywhere
Some conversations an agent can resolve outright, others need a human immediately. The dashboard needs to make that difference visible at a glance rather than treating every conversation the same.
We stopped treating this as an AI chat bolted onto an existing CRM, and started treating it as two collaborators sharing the same conversation.
The problem we kept running into with other tools was familiar: an AI chat interface dropped into a legacy CRM UI, built around the idea that a human operates the AI, not that they work alongside it. Replies still weren't instant, and using the AI well meant learning how to prompt it correctly rather than just working.
We moved away from that. Instead of a human directing an AI through a chat box, WhatsApp Agent Dashboard gives both a real-time view of the same conversation and lets either one act, with restrictions set by the human rather than a workflow the human has to learn. The agent isn't a tool being operated, it's a collaborator with its own visible actions, its own confidence level, and a human who can step in without switching modes.
This is concept work, so the next step is testing the parts most likely to break against a real deployment: whether "Always allow" gets used the way it's intended or whether people default to it out of fatigue and hand over more authority than they meant to, how the confidence score holds up against real conversations rather than becoming a number people learn to ignore, what happens when two agents need to hand a single conversation between each other, and real WhatsApp Business API behavior around message templates and response windows, which the concept doesn't yet account for.
Built to collaborate. Not tested on a real inbox yet.