Skip to main content
Affinity AI Chat is an intelligence layer inside Affinity — like a colleague who has read every company, person, note, and interaction in your CRM. It answers questions grounded in your data with cited sources, and can take a defined set of confirm-before-execute actions, always with your approval.

What it can answer

Affinity AI draws only on data you can already see:
  • Companies, people, and opportunities — fields, lists, owners, stages
  • Notes, files, meetings, emails, and activity history — including notes and files on opportunities and people, and PDFs such as pitch decks already in Affinity
  • Relationships and connection strength
  • Recent external news on a company
A few real prompts: “Help me prep for my meeting with Acme Corp tomorrow,” “Find Series B medtech companies we haven’t contacted in 90 days,” “Who at our firm has the strongest relationship with Sequoia?” Ask follow-ups in the same thread to refine.

What it can do (with your approval)

Every action is confirm-before-execute: Affinity AI proposes the change, and nothing is written until you approve. The actions:
  • Update a field — a field value or an opportunity’s pipeline stage.
  • Add to a list — add one record to one of your lists, or several companies or people in one step.
  • Create a note — draft and log a note on a record.
  • Create a reminder — so a follow-up doesn’t slip.
Chat can also run the Data Update agent on a note: ask “Suggest field updates from my latest note” and review the Suggested Field Updates card that appears (see Meeting Prep & Data Update). Warm Intro runs from Chat too: ask for it whenever you need it. Meeting Prep doesn’t run from Chat, and Chat can’t read your meeting briefs: briefs are made overnight or when you click Generate Brief in Meetings. AI Chat proposing a reminder for Northwind Logistics with a due date and Create reminder Not available in Chat: uploading a file, creating a saved view, creating a list. Affinity AI will not send email or outreach, merge or delete records, or make bulk changes, apart from adding several records to a list in one step (Affinity MCP can delete and merge, with your confirmation). See Take actions with AI Chat.

How an answer looks

Affinity AI works in two visible stages: a thinking state shows the work in progress; a short “sources used” line at the end tells you exactly which data the answer drew on, so you can open it and verify. AI Chat answering who at the firm knows a company best, with relationship strength scores

How to trust it

  • It shows its sources. Every answer ends with the data it reviewed.
  • It signals confidence. When supporting data is thin, it flags low confidence — open the source before relying on it.
  • It suggests; it doesn’t act on its own. You always see a confirmation before anything changes, and can tell a suggested action from a completed one.
  • It respects your permissions. It can only see and change what your account already allows.
  • You can give feedback. 👍 / 👎 on any answer; 👎 routes to Affinity’s product team.

Known limits

  • ~90-day activity-history window — very long-range “first ever interaction” questions can be truncated.
  • People search is by name or email — not by role/title, location, or expertise.
  • No location/geographic filtering.
  • No reliable historical field-value lookup — it reports a field’s value now, not what it was at a past point.

Where to find it

AI Chat lives in the web app (not on mobile). Open it from the left navigation — Chat and Ascend (agents) are separate items — or press Cmd+J (Mac) / Alt+J (Windows). Admins control whether AI features are on for your org at Settings → Automations & AI → Affinity AI.

Common questions

  • Will it change my data without asking? No — every action is confirm-before-execute. There’s no “always allow.”
  • Does it see data I can’t access? No — it’s scoped to your existing permissions.
  • Will it send emails? No.
  • How is it different from Search or filters? It’s conversational — ask follow-ups and refine — and it aggregates across profiles instead of tab-hopping.

Affinity AI can make mistakes. Review each suggestion before approving it.