Supervised AI operations

Run every WhatsApp Sales & CS account from one dashboard.

AI replies automatically, Sentinel reviews every risky message before it sends, and your team can take over anytime.

See how it works

Multi-account. Human-in-the-loop. Every decision logged.

Sentinel Control Center

Live

RRina

Hot

Hi, if I pay the deposit today can it ship tomorrow?

Yes. A 30% deposit locks tomorrow’s shipping slot. I’ll send the details.

Sentinel reviewing replyApproved · low risk

5
AI modes per conversation: 100%
of decisions logged and auditable: 2
control layers: AI plus Sentinel: ID/EN+
languages supported, extensible

Works with the tools you already use

A powerful AI engine, kept honest and in control.

AI that answers from your knowledge, not guesses, in the customer’s language, on a model you choose.

It never makes things up: The AI answers only from your knowledge base. When it does not know, it checks with an admin instead of guessing.
Run the AI model you choose: Works with any OpenAI-compatible endpoint. Defaults to Nous Research Hermes; swap models in config, no code change.
Replies in the customer’s language: Indonesian, English, and more. Set it per bot and the prompts adapt automatically.
Reads the tone, stays on persona: Detects customer sentiment and answers in your brand’s voice, backed by hybrid knowledge retrieval.

Sentinel Supervisor

A control layer above every bot.

Sentinel reads conversation context, scores confidence and risk, then decides: send, hold as a draft, or ask an admin to take over. Deterministic rules catch sensitive keywords, the LLM judges the rest.

Confidence and risk score per reply
Pre-send gate for sensitive chats
Post-send audit for routine chats

Sentinel reviewApprove

AI draft

Yes. A 30% deposit locks tomorrow’s shipping slot. I’ll send the details.

Confidence92

Risk8Low

Reason

Matches the pricing and shipping info in the knowledge base. No sensitive data.

One workspace

Everything a sales and CS team needs.

Multi-account inbox

Every WhatsApp account in one WhatsApp Web-style queue, with live status.

Automatic lead scoring

Cold to very hot, triggered by price, stock, and deposit signals.

Knowledge base with RAG

Edit knowledge and AI answers change instantly. No fabricated data.

Controlled campaigns

Approval, per-account rate limits, human-like delays, and opt-out exclusion.

Closing analytics

Conversion funnel, per-bot attribution, and win/loss breakdown.

Full audit trail

Every send, takeover, and Sentinel decision is logged and exportable.

Conversion by stageClosed this week

New100%

Warm72%

Hot48%

Closing31%

Won22%

Measurable results

See what actually closes.

Conversion funnel, per-bot attribution, and win/loss in one view. Every stage transition is logged automatically.

Conversion at each stage, from new lead to closing
Which bots and personas close the most deals
Why deals are won and lost, with a daily trend

Live in three steps

Connect, train, supervise.

01
Connect accounts
Scan the Baileys QR. Sessions persist, reconnect automatically, anti-ban send delays.
02
Train the persona
Set the Soul/persona and product knowledge base. AI answers on-brand.
03
Supervise and scale
Pick an AI mode per conversation and let Sentinel guard quality.

Per-conversation control

Five modes, from fully automatic to manual.

AI ON

Bot replies automatically, Sentinel audits after sending.

AI SUPERVISED

Sentinel reviews before the message goes out.

AI DRAFT

Bot drafts the reply, an admin sends it.

AI PAUSED

The bot stops automatically when risk is detected.

AI OFF

Every reply is handled manually by the team.

The questions teams ask first.

Will my WhatsApp accounts stay safe from bans?

Every send uses human-like delays and per-account rate limits, sessions persist with auto-reconnect, and there are no broadcast-style blasts. Campaigns always run through a controlled, approved queue.

What if the AI sends the wrong thing to a customer?

For sensitive chats, Sentinel reviews the reply before it sends and can hold it as a draft, block it, or hand off to an admin. When risk is detected, the AI pauses automatically.

Which AI model does it use, and can I change it?

The AI engine works with any OpenAI-compatible endpoint. It defaults to Nous Research Hermes, and you can switch to another model through configuration with no code change.

Does it support multiple admins with different roles?

Yes. There are owner, supervisor, admin, and viewer roles with distinct permissions, plus an audit trail for every action.

Be one of the first teams on Sentinel

Ready to take control of your WhatsApp operation?

We are new, so we will prove it on your own accounts, not with customer logos we do not have yet.