One engine, four jobs
We pick the shape that matches how the business actually sells or supports.
Instead of scoring leads, the bot classifies incoming messages by department, category, priority, and sentiment. A fully-answered question doesn't turn into a ticket. Only what's actually unresolved does.
A ticket queue by department and priority, not a lead list.
- Grocery markethandles order complaints, product-quality issues, and delivery problems
My delivery was supposed to arrive an hour ago and I still don't have it.
Sorry about that. I've flagged this as urgent for the delivery team, they'll message you directly. Can you confirm your order number?
The dashboard for this setup
Every archetype starts here
What every bot can do, regardless of which job it's set up for.
Reads, listens, and looks
Understands text, voice notes, images, and PDFs in the same conversation: a price-list screenshot, a voice note, a product photo, a spec sheet, whatever a customer actually sends.
Voice notes transcribed carefully
Transcription is tuned specifically to avoid mis-detecting a customer's language or dialect, a deliberate engineering choice, not a default setting.
One reply, not three
A purpose-built batching engine folds a burst of two to four messages into a single reply, detects when someone's getting impatient and answers early, and queues follow-ups if it's already mid-reply.
Remembers for real
Contact memory is stored in Postgres, not a session cache, so the bot doesn't re-ask a name, a preference, or a stated concern it already has on file.
Doesn't lose messages
If the bot process restarts mid-conversation, pending messages pick back up instead of vanishing. Retried webhook deliveries are deduplicated, and stickers, reactions, and location pins are handled appropriately instead of treated as new questions.
Official API, dedicated infrastructure
Runs on the WhatsApp Cloud API Meta issues directly to the business, not an unofficial connection that can be cut off. Every client gets their own bot process and their own database.
The dashboard
Built like real internal software, because it has to be.
One inbox for everything
Full chat history, contact profile, and, depending on the archetype, lead score, order status, or ticket state, all in one place.
A person can take over anytime
Any conversation can be picked up mid-thread. The bot steps aside automatically, and in at least one deployment, hands back over on its own if the human doesn't act within a timeout.
Role-based access, audited
Admin, manager, and sales-rep tiers, audit logging, and login rate-limiting. Not a demo dashboard bolted onto a chatbot.
Outreach tooling built in
Bulk-send approved WhatsApp templates from a CSV list, with duplicate-send warnings and per-variable mapping.
Real-time alerts, claimed and resolved
Ahmed asked about bulk pricing on 40 units. Unclaimed.
$1,240 order placed. Claimed by Sara, 2 minutes ago.
Built with a self-serve future in mind, not sold as one today.
Two deployments already meter usage: every paid AI call, a reply, a scoring run, a manual re-analysis, a transcription, is priced from real token counts and tracked against a plan allowance, with soft-stop behavior when a limit is hit. That's not live self-serve billing yet. It's the groundwork for it.
Ready to see this running for your business?
Book a call and we'll map your setup to the archetype that fits.