An automatic check compares replies with what your AI employee did
For business owners and managers
When your AI employee tells you in its chat that it did something or will do it later, that is checked against what it really did. If the check cannot finish, it now tries again, so fewer replies go unchecked.
Why it matters
A note can show you when what your AI employee says and what it did do not match. If it says an email went out, or promises a reminder that nothing backs up, the note tells you, so you can catch a missed step early.
- Before
- If the check was busy or ran out of time, the reply stayed unchecked, with no second try.
- Now
- A reply check that cannot finish tries again 2 minutes and then 15 minutes later, up to three tries in all, and a later note is added just like an early one.
What you get
- What is checked: statements that it did something, such as sending an email, and promises to do something later are compared with its actions, its email and the scheduled tasks it set up.
- Claims about other people: a statement that a client received, read or accepted something is treated as unconfirmed unless a message from that client shows it.
- A short note, never a rewrite: if a statement does not hold, a short note in the reply's language is added at its end, saying what can and cannot be confirmed. The reply itself stays as it was.
- Promises need a plan: a promise to do something later counts only when a scheduled task or an email watch for it is set up. If nothing is, the note says so and tells you to ask your AI employee to set it up.
- No wait for you: the check runs after the reply is shown, so the reply is never delayed. If it cannot finish, it tries again 2 minutes and then 15 minutes later, up to three tries in all.
- An automatic check: it covers replies in your chat with your AI employee. It can miss things, and a reply can still stay unchecked when all three tries fail.
How to use it
Work with your AI employee as usual
Ask it to do something in your chat, such as sending an email or reminding you later. There is nothing to switch on.
Read the reply
The reply arrives as always. The check runs right after it, so a note usually appears within a minute, or up to about 25 minutes later when the check has to try again.
Act on a note
If a note says a promised follow-up is not set up, answer in the chat to have it set up. If it says an action is not shown, ask your AI employee to check it or do it again.
Updates to this feature
Notes from VEP's automatic check are clearly marked
Notes of the check sit in a small muted block under the answer, signed as the VEP check, and every new note ends with one short next step.
Fewer false alarms when your AI team reports on its tasks
The reply check reads each task's state and history, so a report that matches is backed by the task, and one that does not gets a note saying what the task shows.