Skip to content
AptiroAptirodocs
Menu: Connect an AI assistant

Integrations

Connect an AI assistant

On this page

Aptiro can be reached from Claude or ChatGPT, so you can ask about your accounts, contacts, deals, pipeline and activity in plain language without opening the app. By default it is read-only: an assistant can look things up but never change or delete anything. An owner or admin can also let it log activities on your behalf — see "Logging activities" below.

Setting up

An owner sets up each assistant once, from AI Assistant in the sidebar. Setting up Claude or ChatGPT creates the connection for the whole organisation; every person still signs in with their own Aptiro account when they connect it on their own side, and sees only what they can already see in Aptiro. Once it is set up, share the page with your team: it shows each vendor's own connection steps, including the connector URL and client ID to paste in.

The AI Assistant page, with example questions to ask and Claude already set up, ChatGPT and Claude Code not yet
The AI Assistant page, where an owner connects Claude or ChatGPT for the whole organisation

A connected assistant shows Connected on its card, with a tick. An assistant that has not been connected yet says Not set up. Under a connected assistant, the page names who connected it and when, and either the date you last used it or that you have not signed in from it yet.

ChatGPT may issue its own callback URL per custom connector. Sign-in only works once that URL has been added on the AI Assistant page, so if a colleague's ChatGPT connection fails on first sign-in, check whether its callback URL is listed there yet.

Disconnecting an assistant signs everyone in the organisation out of it within a minute. Nothing already in Aptiro is changed, and setting it up again is the same short process.

What it can see

Each request runs with the permissions of whoever is signed in to the assistant, enforced the same way as a request from the app itself: a member with view-only access to deals gets view-only answers about deals, whatever they ask.

Anyone who can manage settings and the team can also limit what an assistant may do for a person or a team below their own permissions, from Settings > AI access. A limit is itself a role, so the assistant gets whatever the person's own permissions and that role both allow, never more than either. This never affects how that person signs in to Aptiro directly, only what an assistant acting for them can reach.

The page lists everyone with what an assistant can do for them today and where that comes from: their own setting, or the teams whose limits apply (or "None of their teams has a limit"). Where several roles limit someone, it reads "Limited to both Viewer and Sales rep" for two roles, and "Limited to all of Viewer, Sales rep, Analyst" for three or more. Each person has one menu:

  • Use their teams' limit: the default. If a team they are in has a limit, it applies; if several do, the assistant gets only what all of those roles allow.
  • No limit: the assistant can do whatever the person can, whatever their teams are set to.
  • Limited to a role: the assistant gets only what the person and that role both allow.

Team limits are set under By team on the same page. The switch that lets assistants log activities has its own Logging activities card below.

Logging activities

An owner or admin can let a connected assistant log activities — calls, emails, meetings, notes — as well as read your CRM, from Settings > AI access. It is off by default: turning it on is a deliberate choice, not something that happens by connecting an assistant.

A person still needs permission to create activities for it to work: the setting widens what an assistant may attempt, never what the signed-in person is allowed. An activity an assistant logs shows "via AI assistant" on the timeline, so a manager can always tell what a person did from what a model recorded; on the Scale plan, the audit log also records which connector did it.

Asking an assistant to log the same thing twice, or an assistant retrying a call that timed out, does not create a second activity: a near-identical entry for the same account within a few minutes returns the one already logged.

If you turn this on and nothing seems to happen: a connection set up before this feature shipped keeps its old, read-only sign-in until it is reconnected — a refreshed sign-in keeps whatever was granted when it first connected, so no amount of waiting changes it. The AI access page names any connected assistant that still needs reconnecting; disconnecting it and setting it up again from the AI Assistant page is what picks up the new permission.

Leader routines

Two ready-made prompts help a manager prep from Claude (or another assistant that supports MCP prompts). Each one tells the assistant which Aptiro tools to call; it still only sees what your role can already see in Aptiro, and it never changes data. Commission and pay figures stay out of the assistant: open Commission in Aptiro when you need attainment.

Prepare a QBR takes a person or team and a quarter. It pulls won, lost and open pipeline for that window, names the top accounts and their open signals, and drafts talking points.

Pressure-test a commit takes a person and a period. It lists the open deals they expect to close in that window, with stage moves, close-date slips, last activity, account signals and gaps (no next step, no activity in a fortnight, close date pushed more than once).

Example asks:

  • "Prep my QBR for Alex Rivera for 2026-Q4"
  • "Prepare a QBR for the North team for this quarter"
  • "Pressure-test Sam Lee's commit for this month"
  • "Pressure-test the commit for member:user_abc from 2026-10-01 to 2026-12-31"

Last updated 2026-11-09

Search the documentation