Control and safety
The questions worth asking are about boundaries.
If you are letting software near your customer conversations, feature lists are not the thing to read. What it will not do matters more than what it can.
What holds, every time
Each one is in place today. If a control is not in place yet, it is not on this list.
A person approves every reply
The AI drafts; someone on your team sends. There is no setting that hands the conversation over, and there is no mode where a message reaches your customer without a human having agreed to it.
Your rules are enforced on the draft
The limits you set (no invented discounts, no promised delivery dates, no commitments you have not authorised, no medical decisions) are applied before a draft reaches the inbox, not left to the AI's judgement in the moment.
It says when it does not know
If an answer is not in what you taught it, it does not fill the gap with something plausible. It hands the conversation to a person instead, which is the correct answer to a question your business has not answered yet.
Your data is only yours
Your conversations, your customers and your business knowledge are kept separate from every other customer's. An agency running several businesses gets the same separation between each of them.
Never used to train AI models
Your customer conversations are not used to train AI models, ours or anyone else's. We do not sell, rent or monetise your data.
On your own official number
ConnectFlow connects to your business number through Meta's official WhatsApp Cloud API. The number and the WhatsApp Business account stay yours, and you can disconnect them.
Every decision is on the record
Who approved a reply, who rejected one, when it was sent and whether it was delivered are all kept. When someone asks what was said to a customer and who agreed to it, there is an answer.
Run with care behind the scenes
Access to the systems that run ConnectFlow is limited to the named people who need it. Your data is backed up, and restoring from those backups is tested rather than assumed.
Not tied to one AI company
ConnectFlow is not built around a single AI vendor. The engine can work with different AI models, and every control on this page applies whichever one is drafting the reply.
Including the ones that are not flattering
A vendor who answers only the easy questions has told you something about the hard ones.
- Can you read our messages?
- The platform reads and stores message content, because that is what makes a shared inbox and a useful draft possible at all. We will not claim otherwise, and we treat what we hold under the commitments in our privacy policy. Access by our staff is limited to what supporting you requires.
- What happens if the AI gets it wrong?
- Then your team sees a wrong draft and does not send it. That is the whole point of the approval step: a mistake costs someone ten seconds rather than costing you a customer.
- Who can see what, inside our own team?
- Everyone you give access to can work the shared inbox for the business they belong to. Nobody sees a business they were not added to.
- Can we get our data out, or delete it?
- Yes. Ask us and we will export it or delete it. The detail of how long we keep what, and on what legal basis, is in the privacy policy rather than summarised loosely here.
The full commitments are written down.
This page is the summary. The privacy policy and the terms are the version you can hold us to, and they are written to be read rather than to be survived.