Skip to main content
The Assistant answers questions about your customers across every account, meeting, request and conversation in Zudo, and helps you act on the answer. Ask it the way you’d ask a colleague: “Which of my accounts need attention this week?”, “How much MRR does each CSM own?”, “Draft follow-ups for today’s meetings and create tasks for what we promised.” It looks things up in your data as it goes, shows you each lookup, and answers with links to the records it used. When you ask it to change something, it proposes the change for you to review. Nothing changes until you apply it.

​
Where to find it

  • Assistant in the main navigation opens a full page with your chats down the side. Start a New chat for each topic.
  • Ask the Assistant on an account, meeting or request opens it in a panel about that record. Questions that don’t name anything else are about the record, and the panel offers quick actions written for it, such as Prep my next meeting on an account or Suggest follow-up tasks on a meeting.
Your chats are private to you. Nobody else in your organization can open them, admins included.

​
What it can read

The Assistant reads Zudo through the same operations as the API, as you. It sees exactly what you can see and nothing more. That covers:
  • accounts with their health, MRR, CSM, lifecycle status, traits, segments and indicators
  • contacts, and your team’s members
  • flags, requests and their linked issues
  • meetings, with transcripts and notes
  • email and Slack threads
  • tasks and notes
  • NPS responses
  • product usage events and daily rollups
  • AI briefs already generated in Zudo, such as Account Summary and Meeting Prep
For counts and totals across many accounts (“how many at-risk accounts does each CSM have?”) it asks for the totals directly, without reading every account. It leaves churned accounts out unless you ask about them.

​
Making changes

Ask the Assistant to create or update something and it proposes the change on a card under its answer. The card lists every field, with accounts, people and meetings shown by name. Check the changes you want and choose Apply, or Dismiss them. Changes are made as you, so they follow your permissions, and the Assistant is told on your next question which ones you applied. It can propose:
  • creating and updating tasks
  • adding notes to accounts, meetings, requests and issues
  • updating account fields, such as the CSM, and your own traits
  • creating and updating custom object records
  • dismissing or snoozing flags
  • creating and updating requests, and linking issues to them
  • creating and updating contacts, and adding meeting attendees
It can also propose changes to how Zudo watches your accounts:
  • Segments. It writes the conditions from the same fields the segment builder offers, and the card says how many accounts the segment matches today and names a few.
  • Flag rules, which raise a risk or opportunity flag on every account in a segment. When no segment fits, it proposes the rule with a new segment, and applying it creates both or neither. The card says how many flags the rule would raise today.
  • Indicators from your product events. The card says how many accounts the indicator holds for today.
  • Playbooks, with a trigger and steps: tasks, flags, segment changes, trait updates, Slack messages, NPS surveys, waits and branches. The card lists every step in order and says how many accounts the trigger matches today. A new playbook is created switched off, and changes to an existing one are saved to its draft, so nothing runs until someone turns it on or publishes it in the playbook editor. It can’t add or change webhook steps.
So “flag accounts at risk when report exports drop” becomes one proposal: a flag rule on a new segment of accounts whose exports fell by half against 30 days earlier. These follow the same permissions as the screens they belong to: only admins can create or change segments and indicators, and anyone can manage flag rules and playbooks. They’re only available in the Assistant, not through the API or MCP, so a person always reviews them first. It can’t send emails or messages. Drafts appear in the answer with a Copy button. It also can’t create accounts, switch a playbook on or publish one, or delete anything.

​
Work across several records

For the same job across several records, such as a follow-up for each of today’s meetings, the Assistant splits the work into up to 10 sub-tasks that run side by side. Each sub-task shows its progress, and you can open it to read its answer. Changes the sub-tasks propose all land on the one review card.

​
Searching the web

The Assistant can search the web for news about a customer or their industry, and read a page you link to. It says where what it found came from. Admins can turn web search and opening pages off in Settings → Organization → AI. With both off, it only uses your data in Zudo.

​
Memories

Memories are short notes the Assistant keeps in mind in every chat they apply to.
  • Organization memories apply to every chat in your organization: your terminology, how to read your data, the tone you want. Only admins can add or change them. Open Memories on the Assistant page.
  • Account memories apply to chats about one account: who decides, how they work, what they call things. Anyone can add them from Memories in the account’s Assistant panel.
Keep each memory short and stable. The Assistant reads every memory on every question it applies to, so long documents and figures that change week to week don’t belong here.

​
Prompts

Type / in the chat box, or choose the sparkle button beside it, to pick a prompt:
  • Your saved prompts are yours alone. Choose Save current to save what’s in the chat box, or ask the Assistant to save a prompt for you.
  • Zudo’s prompt library has ready-made prompts for planning the week, churn risk, meetings and follow-up, renewals and expansion, QBRs and team operations. A record’s panel shows the ones written for that kind of record.

​
Usage limits

On Zudo’s AI, each person has two usage limits: one over the last 5 hours and one over the last 7 days. The percentage under the chat box shows the fuller of the two, and clicking it shows both. Questions that read more records, run sub-tasks or search the web use more of the limit than short ones. When a limit is full, the Assistant says when you can ask again. Admins can see how much of the weekly limit each member has used in Settings → Organization → AI. To remove the limits, add your organization’s own key.

​
Long chats

Each answer rereads the whole chat, so a long chat answers more slowly and less well. The Assistant warns you when a chat is getting long, and once a chat is full it asks you to start a new one. Starting a New chat for each new topic keeps answers sharp.

​
Use your own AI key

By default the Assistant runs on Zudo’s AI, with the usage limits above. An admin can add your organization’s own OpenAI or Anthropic API key instead, in Settings → Organization → AI → Use your own AI key:
  1. Choose the provider and the model. OpenAI offers GPT-6.1 Sol and GPT-6 Luna; Anthropic offers Claude Opus 5.5, Claude Sonnet 5.5 and Claude Haiku 5.5.
  2. Paste the API key and choose Add key.
Zudo checks the key with the provider before saving it, stores it encrypted, and never shows it again. Admins see only its last four characters. With your own key:
  • the usage limits don’t apply
  • your provider bills you for what the Assistant uses
  • your questions and the data the Assistant reads go through your own account with that provider, under your agreement with them
  • the key applies to the Assistant only. Other AI features, such as Account Summary and AI Pulse, keep running on Zudo’s AI.
To go back to Zudo’s AI, choose Remove on the saved key.

​
Good to know

  • Answers come only from your data in Zudo, plus the web when that’s on, and can be wrong. Check before acting on one.
  • Stop ends an answer partway. What it found so far is kept.
  • Dates are in your own timezone, and “my accounts” means the accounts you’re the CSM of.
  • For a quick question about one account with citations to its emails and meetings, Ask AI on the account’s Search tab is still there.