Put AI in front of customers without losing the wheel.
Every control below is on, in the product, today. Human review where you want it, automatic escalation when a conversation turns, guardrails per agent, and a full audit trail of what was sent and by whom.
Controls that are on, in the product, today.
- Opt-out suppression
- An opt-out is recorded against the contact and suppresses future automated sends to that person on SMS, WhatsApp and voice, not just the channel they replied on.
- Human-in-the-loop review
- Queue AI-drafted messages for a person to approve before they send. Always on, or triggered only when sentiment or risk signals trip.
- Sentiment escalation
- A conversation that turns is handed to a human automatically, with the full history attached rather than a summary.
- Per-agent guardrails
- Each agent carries its own instructions and limits, so what one is allowed to say does not silently become what all of them can say.
- Grounded answers
- Agents reason from the knowledge base you publish, with staged publishing so a change is reviewed before it reaches a live conversation.
- Office hours
- Campaigns schedule sends inside the hours you set for the team, instead of firing at 3am in the contact’s time zone.
- Per-region AI voice switch
- Turn synthetic voice off for a market and run it fully human, while agents keep working SMS, email and chat there.
- Sender identity
- Sender ID and business identification are configurable per region, along with the opt-out wording that goes out with your messages.
- Audit trail
- Every message and call records what was sent, by which agent or which person, and when. Inspectable per conversation, months later.
- Role-based access
- Admin and operator roles per workspace, with separate workspaces for teams that need harder separation between brands or clients.
- Your own carrier
- Telephony runs through your account with Twilio, Telnyx, Vonage, Plivo or Sinch, so your numbers and sending reputation stay yours.
Turn it up as fast as you trust it.
Most teams start with every outbound message queued for approval, watch what the agents draft for a week, then hand over channel by channel. By the second week the approval queue is usually a formality.
You hold the dial the whole time. Tighten it for a regulated market or a list you have not touched in a year, open it up where the agents have already proven themselves, and change your mind on a Tuesday without opening a support ticket.
The questions procurement asks.
Reporting a vulnerability instead? The security page has the scope and the rewards.
Do you train models on our data?
No. Your customer data, conversations and pipeline are not used to train models. Agents are grounded in the knowledge base you publish so that answers come from your material rather than from anything absorbed from another customer.
Can we require a human to approve every AI message?
Yes. Human-in-the-loop review can be on for everything, or scoped to trigger only when sentiment or a risk signal fires. Drafts queue in the shared inbox for a person to edit, approve or discard, and a rep can take any live conversation off an agent mid-thread and hand it back afterwards.
Does RevOps.ai keep us compliant with TCPA, GDPR or local messaging law?
We give you the control surface, and it is a deep one: opt-out suppression on automated sends, human review before anything goes out, configurable sender identity per region, office-hours scheduling, per-region AI voice switches, and a full audit trail. Because compliance also depends on your list, your basis for contacting it and the markets you operate in, teams normally set these controls with their own counsel and we are happy to work through that with you.
What is actually logged?
Every message and call carries what was sent, which agent or which person sent it, and when, retained per conversation and inspectable later. Calls additionally record duration and outcome, and a recording where you have enabled it. Recording is a setting rather than a default, because recording and disclosure law differ by jurisdiction.
How do we separate client data as an agency?
Separate workspaces per client, each with its own agents, numbers, contacts and members, managed from one dashboard. Roles are per workspace, so someone working on one client account is not implicitly on another.
Where do we report a security vulnerability?
On the security page, which sets out scope, what is out of scope and the reward for each severity. That page is for vulnerability disclosure. This page is about how the product behaves when it is working correctly.