Uma, the assistant
Uma, your team’s assistant
Ask the way you’d ask a teammate and get answers from what’s in your meruma. Ask for a change and it gets prepared for you to confirm. Included in every plan.
Sample conversation with Uma. Question: “Which chats are waiting for a reply?”. Uma: “Three. The one waiting longest is Emma Collins: she messaged you on WhatsApp 2 hours ago to set up the visit”. Uma suggests reminding you to call Emma today at 5:00 pm; once you confirm, she replies: “Done. I’ll remind you at 5:00 pm”.
Uma, the assistant
What changes
-
Answers from what’s in your meruma
“What’s on today?”, “which chats are waiting for a reply?”, “how are sales doing this month?”. It searches and summarizes contacts, cards, chats, your calendar and reports, using your organization’s data.
-
Explains how things work
It answers “how do I…?” with the help center articles and leaves you the link. If there’s no article on it, it says so instead of making up steps.
-
Prepares the change, you confirm it
A reminder, creating or moving a card, assigning contacts, leaving a note or messaging a customer. Uma shows you a card with what it’s about to do: you confirm or cancel, and almost everything can be undone.
-
Sees what you see, nothing more
It works with your permissions: whatever your role can’t see, it can’t either. It doesn’t read the fields your organization marks as sensitive, and your conversations with Uma aren’t used to train models.
What people usually ask
Does Uma cost extra?
No. It’s included in every plan and doesn’t use any key of yours. There’s a usage limit per person per day and another per organization per month, so there’s enough for everyone.
How is it different from the AI agents?
AI agents answer your customers in the chats. Uma works for your team: it helps you and your teammates inside meruma.
Can it message my customers?
Yes, by chat and always with your confirmation: the card shows the exact text, the channel and who it goes to. A sent message can’t be undone; if you’d rather review it, ask for a draft.
Which model does it use, and what happens to my data?
Uma uses Anthropic’s Claude models on Amazon Bedrock. Your conversations are yours: nobody else in the organization can read them, they delete themselves 90 days after the last message and they aren’t used to train models.