AI control
A robot that follows your rules, and explains itself
Give the robot rules: never promise a refund, never quote a price that isn’t on the price list, never offer a bigger discount than you allow. Every reply is checked against the rules before it goes out, each reply comes with an explanation of what the robot relied on, and a manager who wants something changed just writes it a sentence.
Capabilities marked “Soon” on this page are being built now and will be available to everyone later. This is how they’ll work once they ship. Everything new
What robots do todayWhy it matters
-
Peace of mind
Every reply is checked against your rules before it goes out, and one that breaks a rule is rephrased or handed to a rep.
-
Know why it said that
Next to every robot reply you see the knowledge it found, the tools it ran and the guidance it followed.
-
Fix it in a sentence
A manager writes “don’t offer a discount before asking about the budget”, and the robot confirms how it will act next time.
What’s inside
Capabilities
- Soon
Rules the robot won’t cross
Set rules for each robot: topics it must not discuss, promises it must not make, and prices and discounts only as you set them. Every reply is checked before it goes out. A reply that breaks a rule is rephrased, and if that doesn’t fix it, the conversation goes to a rep.
- A quick check before every reply, without holding up the conversation
- An extra AI check for rules that depend on context, if you turn it on
- Money actions are checked before they run
- A tester: write a sample reply and see whether it passes
- Soon
Clear disclosure to customers
The robot tells customers it’s an AI assistant, once at the start of the conversation, in the wording you choose. And a customer who asks “am I talking to a robot?” gets a truthful answer: the robot doesn’t present itself as a person.
- A disclosure message once per conversation
- In the wording you choose
- A truthful answer to “is this a robot?”
- A reply in which the robot claims to be human is never sent
- Soon
A monthly trust report
Once a month you get the full picture: how many replies were checked, what was blocked and why, and how many conversations went to a rep. Every block is kept in a feed you can review, and the report downloads as a CSV file.
- What was checked, what was blocked and why
- Conversations handed to a rep
- A feed of blocked replies, to review
- Download as a CSV file
- Soon
Coach the robot in a sentence
On any robot reply, a manager clicks “Coach the robot” and writes what to change: “don’t offer a discount before asking about the budget”. The robot turns it into guidance and confirms what it will do next time, without anyone editing its instructions.
- A “next time I will…” confirmation, in Hebrew or English
- A check that new guidance doesn’t contradict existing guidance
- A history of all guidance, with one-click undo
- Every change is recorded in the audit log
- Soon
“Why did it say that?”
Next to every robot reply you can open an explanation: what the customer wrote, which knowledge was found, which tools the robot used, which guidance it applied and why it decided to answer or hand over.
- The knowledge found, with its source
- The tools the robot ran, with secrets masked
- The guidance it applied
- Only what the viewer is allowed to see
How it works
From setup to results
- 1
Set the rules
Topics, promises and prices the robot won’t cross.
- 2
The robot replies
And every reply is checked against the rules before it goes out.
- 3
See why
The explanation next to each reply shows what the robot relied on.
- 4
Fix it in a sentence
New guidance, and the robot confirms what will change.
Control, safeguards and privacy
These checks work alongside the tool approvals and rep handover you already have.
- Rules and guidance are set by managers
- Blocked replies are stored encrypted
- The explanation never shows secrets, or conversations the viewer isn’t allowed to see
- Any guidance can be undone, bringing back the previous version
- The extra AI check stays off until you turn it on
More in MoreSend
- Campaigns Every customer gets the message on the channel they actually read, at an hour that suits them, and only if they agreed to hear from you.
- Leads and growth Soon A comment on a post becomes a private conversation, an ad lead gets a WhatsApp within a minute, your site chat reaches out at the right moment, happy customers bring friends, and you see the revenue each conversation brought.
- Smart inbox Soon Live translation in the chat, a ⌘K command bar and macros, snooze until the customer replies, internal threads and side conversations, the next step in one click, and an automatic wrap-up to the lead record.
- Voice Soon AI voice agents that know the customer, WhatsApp calls kept in the same chat, “Call me now” from any chat, WhatsApp voice notes the robot understands, reps who answer in the browser, and a summary of every call on the customer record.
- Studio Soon Build a flow from a plain-language description or start from a ready-made pack for your industry, test it on AI customers and real past conversations, and roll it out gradually.
- Robot knowledge Soon The robot learns your whole website, your files and what’s written in Google Drive, Notion, SharePoint and Confluence, stays up to date by itself, and shows where each answer came from.
- AI assistant Soon An assistant that knows which screen you’re on, answers from the help centre, guides you on the screen itself and acts only with your approval.
- Commerce and payments Soon Sell inside the conversation: product cards, a cart, payment by secure link, quotes signed on the phone, and a personal page where customers see their quotes and payments.
- Bookings Soon A built-in calendar: the robot books, moves and cancels appointments on WhatsApp and by phone, sends a reminder with “1” to confirm and “2” to change, syncs with Google and Outlook, and never books two customers into the same slot.
- Team operations Soon Every conversation reaches the right rep, service targets and shifts that know the Israeli calendar, a live board for the office screen, and QA that turns into coaching.
- API and MCP Soon Your AI assistant or IDE works with MoreSend through an MCP server: find a customer, read a conversation summary, create a lead and send a message, with scoped API keys.
- Agencies and partners Soon One console for every client you manage: client workspaces, usage and channel health at a glance, alerts when something breaks, and support access only with the client’s consent.
Want to see how it would look for you?
In a short demo we’ll show you what works today, and how the new capabilities will fit into your work.