> ## Documentation Index
> Fetch the complete documentation index at: https://docs.every.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Prospecting Agent
> An autonomous AI agent that mines your existing network, surfaces warm sales opportunities, drafts outreach in your voice, and learns from your goals and feedback
## Overview
The **Prospecting Agent** is Every's autonomous deal-finder. It scans the network you already have — your contacts, the people you email, who's on your calendar, and anyone you've imported from LinkedIn — scores each potential deal for fit, relationship strength, and timing, drafts a first outreach message in your voice, and drops the strongest candidates into a **Pending** review queue for you to approve.
You stay in control: nothing is sent and no deal enters your pipeline until you approve it. Once you do, the [Sales Pipeline](/prospecting/pipeline) and its agent take over to sustain the conversation.
The Prospecting Agent runs from the Pipeline page, and approved deals land there
Prospecting runs are AI runs and consume credits. Larger scans (more candidate contacts) use more credits. See [Plans & Credits](/billing/payments) for how usage is metered.
## What the agent does
When you start a scan, the agent works through a full discovery pass:
1. **Reads your business context** — what you sell, who you serve, and your active prospecting goal (your "selling motion").
2. **Looks at your existing pipeline** so it doesn't re-surface deals you're already working.
3. **Enriches sparse contacts** — fills in company, role, and other professional details for thin records so they can be evaluated.
4. **Lists eligible network contacts** — everyone who clears a minimum bar of professional information.
5. **Triages and deep-researches** the most promising candidates.
6. **Scores each one** on fit, relationship, and timing.
7. **Drafts an initial outreach email** in your voice for the deals worth pursuing.
8. **Saves the results** into the **Pending** review column.
Discovered deals always start as **Pending** — the agent never places a deal directly into your visible pipeline. You approve or dismiss each one.
## Where it pulls from (your network)
The agent prospects across **your existing relationships**, not a cold-purchased list. It considers contacts from every source you've connected, with no bias toward any one of them:
* **Manually added contacts**
* **Google Contacts** (import and sync)
* **Gmail** — people you correspond with, and recent correspondence (used to gauge relationship and timing)
* **Google Calendar** — meeting history with a contact
* **LinkedIn imports** — connections and message history you've uploaded
* **Booking-derived and enrichment-provided** contacts
A contact needs at least one real professional anchor — a company, a title or role, an industry, a professional link, or a business email domain — before the agent will evaluate it. Name-only or personal-email-only contacts stay below the bar until enrichment finds better clues. For deeper signal, the agent reads recent Gmail correspondence (about 180 days) and calendar meeting history (about a year) for the contacts it shortlists.
The agent reads from the data sources **you've connected** to Every. Connect [Google Workspace](/integrations/google-workspace) and import your [contacts](/clients/contacts) to give it the richest network to work from.
## Scoring
Every candidate the agent shortlists gets a structured score made up of:
* **Fit** — how well the contact matches what you sell and who you target.
* **Relationship** — how warm the connection is (correspondence, meetings, shared history).
* **Timing** — signals that now is a good moment to reach out.
* **Confidence** — how sure the agent is in its read; low-confidence deals are penalized.
These combine into a single total score the agent uses to decide which deals are worth surfacing and drafting outreach for. Each deal carries its evidence and rationale, visible when you open it.
## Setting a prospecting goal (your selling motion)
A **prospecting goal** is a *selling motion* — an angle plus a targeting segment — that tells the agent what to look for. You set it from the prospecting card on the [Pipeline](/prospecting/pipeline) page using a short three-field intake:
1. **What are you trying to sell?** (required) — accepts `@`-mentions of your [services](/billing/services-and-products), which resolve to specific offerings.
2. **Who do you want to target?** (optional) — the segment you're going after.
3. **Anything else?** (optional) — extra context appended to the goal.
If you don't set a goal, the agent runs a sensible **default** derived from your business context — its "What are you selling?" shows your services summary so you can see what it's working from. The default is a fallback identity; it never prompts you to review or merge it.
### Reusing vs. creating goals
When you enter a new goal, Every checks whether it matches a selling motion you already have:
* An **exact match** quietly reuses the existing goal.
* A **likely match** prompts you: *"This looks like your existing goal 'X' — reuse it or create a new one?"*
* A **genuinely new** motion creates a fresh goal and re-opens your full contact pool for it.
This matching is powered by a fast classifier so similar motions ("sell branding to hospitality" vs. "rebrand work for restaurants") collapse into one goal instead of fragmenting your pipeline. Only goals you type are matched — the auto-derived default never triggers a merge prompt.
Different goals scan different slices of your network. A contact you dismissed under one goal becomes eligible again under a different goal, so you can re-prospect the same network from a new angle without starting over.
### Available Contacts
As you set or change a goal, the prospecting card shows an **Available Contacts** count scoped to that goal — how many people in your network are eligible to be scanned for *this* motion. It changes by goal because contacts already approved, dismissed, or recently triaged under the same goal are excluded.
## Running a scan
There are two ways the Prospecting Agent runs.
### On demand
Open the prospecting card on the [Pipeline](/prospecting/pipeline) page, set or confirm your goal, choose a **scan size** (100, 200, 500, or 1,000 candidate contacts), and run it. Larger scans cover more of your network but cost more credits. The agent works through discovery and you can watch the run progress; results land in **Pending** when it finishes. You can also kick off a run by asking [Every Agent](/ai/every-agent) in chat (e.g. *"prospect my network for design clients"*).
Networks larger than your scan size are processed incrementally across runs — never-scanned contacts come first, and a freshness window keeps the agent from re-evaluating people it just looked at.
### As a recurring Scheduled Task
You can put prospecting on autopilot as the built-in **Warm Leads** [Scheduled Task](/scheduled-tasks/scheduled-tasks). It scans your network on a schedule (a daily local cadence by default) and surfaces new Pending deals for you to review — so warm opportunities keep showing up without you starting a scan each time.
Set it up from the prospecting card's schedule pill or from the Scheduled Tasks hub. Scheduled runs bill credits like a manual run, and the goal is re-resolved on each run.
Schedule recurring prospecting and other agent tasks
## Reviewing results
Discovered deals appear in the **Pending** column on the [Pipeline](/prospecting/pipeline) page. For each one you can:
* **Open the deal** to see the score, the evidence the agent found, and the drafted outreach.
* **Approve** it — the deal moves into your **Lead** stage and the [Pipeline](/prospecting/pipeline) agent takes over.
* **Dismiss** it — the agent learns not to surface that contact for this goal again.
No outreach is sent automatically from the Pending queue. You preview and approve first.
## How it learns
The Prospecting Agent improves from what you do. Approving, dismissing, sending outreach, and saving prospecting settings all feed a **learning loop**: the agent reviews your explicit decision and, when the feedback is clear and durable, updates its own guidance so future runs reflect your taste.
Learning is goal-aware. Durable feedback ("never pitch competitors") generalizes across your org; goal-specific feedback stays scoped to that selling motion. Over time, scoring, enrichment, and outreach tone all sharpen to match how *you* prospect.
The most powerful way to teach the agent is to act on its suggestions. Every approve, dismiss, and edited outreach draft is a signal.
## Your professional profile
Prospecting quality depends partly on knowing who *you* are. Every keeps a per-user **professional context** (title, company, location, industry, a short summary, and your professional links) that the agent uses to disambiguate contacts and judge relationship strength. You can review and edit it — and let Every auto-enrich it — from your profile settings.
## Next steps
Approve deals and let the pipeline agent run the sales motion
Schedule recurring prospecting and pipeline reviews
Give the agent your Gmail, Calendar, and Contacts to prospect from
Import and organize the network the agent prospects
**Need help?** Contact support at [hello@every.ai](mailto:hello@every.ai)