Skip to main content

​
What Ask AI is

Ask AI is an internal AI assistant for your team. It sits in the dashboard sidebar (/dashboard/ask-ai) and can answer questions across your entire HaloAgents workspace, then take real actions like drafting a broadcast or importing users from a CSV. It’s not for your customers. It’s for support reps, account managers, engineers, and founders who want a faster way to find an answer or get something done. You can think of it as: a senior teammate who has read every ticket, every CRM note, every meeting transcript, every health score, every Stripe customer record, and every help article, and is happy to dig through it all on demand.

​
Where to find it

In the dashboard sidebar, click Ask AI (the sparkles icon). The page is gated by the ask_ai:use permission, so only members with that permission see it. See Roles & Permissions.

​
What it can do

​
Look up entities

Ask for a person or a company by name, email, domain, or ID and Ask AI pulls the full profile. 
Who is [email protected]?
What's going on with Acme Corp?
Pull up the last 5 conversations with anyone at orbitforms.ai.
It returns plan, health score, MRR, ARR, renewal info, custom fields, recent activity, and links you can click into.

​
Search tickets and conversations

Show me last week's tickets that mentioned SSO.
What's the latest unresolved ticket for Acme?
What conversations did we have about pricing in March?
Ask AI can also do content searches across your knowledge base, integration data, and ticket history. “What did we tell customers about the export feature?” pulls from help articles, meeting transcripts, and prior responses.

​
Run aggregate queries

How many tickets did we get yesterday?
What's our open ticket count by priority?
Top 10 companies by MRR.
What's the health score distribution this month?
When the result is numerical, you can also ask for a chart and Ask AI renders a real bar/line/area/pie chart inline. Try: “Show me a bar chart of new tickets per day this month.”

​
Pull integration data

Ask AI has read access to:
IntegrationWhat you can ask
HubSpot”How many open deals over $5K closing this quarter?” “Show pipeline by stage.” “What was the last call with Acme?”
Stripe”Pull Stripe info for [email protected].” “MRR by plan.” “Failed payments this week.”
Intercom”Most recent Intercom conversation with this user.”
Zoom / Fathom”What was discussed in the last call with Orbit?”
PandaDoc”When does our contract with Acme expire?”
Slack”Most recent Slack thread about the export bug.”

​
Draft automations

Ask AI can also build broadcasts and series for you in one shot. For example: 
Draft a 3-email re-engagement series for Pro users who haven't logged in in 30 days.
Ask AI will:
  1. Run a dry-run audience preview so you can sanity-check the recipient count and a few sample users.
  2. Create a draft series with the steps, subjects, bodies, and delays already filled in.
  3. Hand you back the dashboard URL so you can review and click Activate.
It never auto-sends. Drafts always land in status='draft' and stay there until a human flips the switch. See Automation AI Builder for the deeper builder.

​
Import users from CSV

Drag a CSV into the Ask AI chat input and ask “Import these users.” It maps columns, dedupes against existing users, and creates contacts in HaloAgents. You see a summary of created vs. updated vs. skipped before anything is committed.

​
Browse the web

If a question needs information from outside your workspace (changelog from a vendor, current event, a public doc), Ask AI can run a web search or fetch a specific URL and weave the result into its answer.

​
How to talk to it

Ask AI is meant to feel like Slack. Some patterns that work well:
  • Be specific about who and when. “How many tickets last week?” is better than “How many tickets recently?”
  • Use real names. Treat company and user names as proper nouns. Ask AI looks them up automatically.
  • Ask for a chart when the answer is numbers. “Show me a chart of…” triggers the chart renderer.
  • Iterate. It remembers the conversation. After “Pull up Acme Corp,” you can say “What did we say to them about SSO?” and it knows the entity.
If a question is broad (“review our pipeline”, “why are we losing deals”), Ask AI will ask 1-3 quick clarifying questions before doing the work. The answers come back as clickable chips.

​
Thinking mode

Click the brain icon next to the input to switch on thinking mode. The model is allowed to reason for longer before answering, and you’ll see a collapsed “Thinking” block with the trace. Best for:
  • Multi-step research (e.g. “Compare last quarter and this quarter across renewals, churn, and pipeline”)
  • Open-ended questions
  • Anything where you’d want a human teammate to slow down and think
For quick lookups, leave thinking mode off so answers come back faster.

​
Voice

Press the mic icon to dictate your question. Click the speaker icon on a response to have Ask AI read it back. Useful when you’re in a call and don’t want to type.

​
Sharing chats

Every Ask AI conversation can be shared with another teammate. Click Share in the chat header to copy a link they can open and read. They can fork it into their own chat and continue the conversation. Useful for handing off context (“here’s what I dug up on Acme, take it from here”).

​
What Ask AI can’t do

A few intentional limits:
  • No PII redaction for team members. Your team has legitimate access to the data, so Ask AI returns names, emails, and amounts without masking.
  • No auto-activation. Even when Ask AI drafts a broadcast or series, it never flips the status to active. A human always reviews and activates.
  • No edits to Stripe, HubSpot, or Linear directly from chat. It can read those, but writes go through agent actions or the dashboard. See Agent Actions.
  • No customer-facing answers. Ask AI’s tone, content, and tool access are tuned for internal use. For customer-facing AI, configure a regular agent.

​
Permissions and access

Access is controlled by the ask_ai:use permission. By default:
  • Super Admin and Admin have it.
  • Members can be granted it individually under Settings > Team, or via custom permission presets.
Ask AI itself respects RLS, so a member only sees data they’d be allowed to see in the dashboard. A teammate without tickets:view won’t get ticket lookups even though Ask AI itself is on.

​
Where to go next

AI Agents

Configure customer-facing AI agents (separate from Ask AI).

Automation AI Builder

The deeper drafting flow when you start from the Automation tab.

Roles & Permissions

Grant or revoke ask_ai:use for individual teammates.

API Reference

Programmatic access to the same data Ask AI uses.