Teleperson
Features

AI Company Overview

A Claude-generated firmographic snapshot, cached once across all users.

For any company in the catalog, Teleperson surfaces a clean, structured overview generated by Claude. The interesting part isn't the AI itself — it's the caching layer that makes the feature economically sustainable.

Overview accordion expanded for Delta Air Lines — description, HQ, employee range, revenue range, founded year, and the top reasons people visit

What's in the overview

Every overview includes:

  • A 1–2 paragraph description of the company.
  • Headquarters location.
  • Employee range (e.g. 1,000 – 5,000).
  • Revenue range (e.g. $100M – $500M).
  • Founded year.
  • Top three reasons people visit (e.g. track an order, dispute a charge, cancel a subscription).
  • Three to five competitors / similar companies you can follow with one click.

The overview lives in a collapsible accordion on the Overview tab. Click any field to see Claude's full reasoning.

How caching works

When an overview is requested:

  1. Check the company_ai_overviews table for an existing entry.
  2. If found, return it — no Anthropic call.
  3. If not, call Claude (Haiku-class model), parse the structured output, persist to Supabase, return.

The overview is company-scoped, not user-scoped. Once Teleperson has an overview for delta.com, every user who visits delta.com from then on inherits it. We never pay for the same company twice.

Refreshing

A manual Refresh button regenerates the overview. There is no automatic refresh cadence — overviews are stable enough that paying tokens on a schedule isn't worth it. Admins can force-refresh from the dashboard if a company's firmographic changes (acquisition, rebrand) warrant it.

Why this matters

Consumer AI products get expensive when every user pays tokens for context the company has already produced. Teleperson's design assumption is that most context is shared — the AI Overview, the top reasons people visit, the list of competitors — and only the response to a specific user's question is genuinely personal. We pay tokens once for the shared context and stream it to every user from Postgres.

  • Ask AI → — the AI assistant that uses the cached overview as grounding.
  • Company Hub → — where the overview surfaces.

On this page