# Agents

> Ready-made agents run on schedule and bring the report to you — weekly at-risk deals, 1:1 prep brief, slip pattern, loss reason summary and competitor price log; numbers that open their records, suggested next steps and an optional AI summary.

An agent is a ready-made report that asks a question for you on a regular schedule: *which deals are at risk this week*, *what should I discuss with this rep in our one-to-one*, *why are close dates slipping*, *why are we losing*, *what prices are competitors quoting*. It runs on the day and time you choose and arrives as a notification; **Run** produces it right away.

Agents live under **Automation › Agents** (`/ajanlar`; Automation in the menu, *Agents* in the command palette). They're available on every plan. Reports are rule-based — the figures are calculated directly from the records in the CRM; AI only writes the summary, and its numbers are checked against the report ([below](#ai)).

## Ready-made agents {#templates}

| Agent | What it does | What it looks at | Default schedule |
|---|---|---|---|
| **Weekly at-risk deals** | Open deals with warnings, red health or an approaching close; what entered risk and what recovered since the last report | [Deal warning rules](/en/guides/pipeline-forecast#warnings), health, expected close, the previous report | Every Monday 08:00 |
| **1:1 prep brief** | Before a one-to-one with a rep: period performance, standing against target, at-risk and slipping deals, conversation coaching, overdue tasks and questions to discuss | [Dashboard](/en/guides/dashboard) performance, [Forecast](/en/guides/pipeline-forecast), warnings, close date pushes, [conversation intelligence](/en/guides/conversation-intelligence), overdue tasks | Every Friday 15:00 |
| **Slip pattern** | Deals whose close date was pushed out: who, at which stage, by how much and how often; repeat slippers and the win rate of pushed deals | Opportunity change log, stage history, closes in the last 12 months | First day of every month 09:00 |
| **Loss reason summary** | Lost and canceled deals: recorded reason, competitor, stage lost at, compared with the objections raised in conversations | Close reasons, conversation intelligence objections and competitors, the previous period | First day of every month 09:00 |
| **Competitor price log** | Collects competitor prices mentioned in call recordings and visit notes; those you approve are written to the competitor details on the company card | Conversation intelligence price signals, competitors on company cards, the deal's product | Every Monday 09:00 |

**Set up** in the gallery opens the agent's settings; **Set up and get the first report** saves the agent and produces the first report immediately.

## Agent settings {#settings}

| Field | Description |
|---|---|
| **Name** | The template's name when left empty |
| **Schedule** | *Every week* (day + time) · *Weekdays* (Monday–Friday, time) · *Every month* (day 1–28, time) · *Manual only*. The page shows when the next report is due |
| **Scope** | For admins, *Whole team* or one rep. An agent set up by a sales user always looks only at their own records |
| **Reps** (1:1 brief) | Admins pick the reps to prepare briefs for; a sales user sets it up for themselves |
| **Agent-specific** | At-risk deals: minimum amount, list limit · 1:1: period (Last 30 days · Last 90 days · This month · This quarter) · Slip: lookback, *repeat slippers: at least* · Loss: lookback · Competitor price: lookback, *Save prices without approval* |
| **Who gets the report** | Always the person who set it up. For an admin's agent, also *Send to other admins* and *Send to specific people* |
| **AI summary** | When AI is on in the workspace and for you, the report comes with a short executive summary |
| **On** | An agent that's off doesn't run on schedule; it can still be run by hand |

Each user can set up up to 30 agents; the last 60 reports are kept per agent.

## Report page {#run}

Every run is a report (`/ajanlar/calistirma/<id>`):

- **Indicators** — depending on the agent: *At-risk deals*, *Amount at risk*, *Closing in Q4* (the current quarter), *Newly at risk* · *Reps*, *Won*, *At-risk deals*, *Overdue tasks* · *Slipped deals*, *Repeat slippers*, *Average push*, *In open deals* · *Lost*, *Lost amount*, *Reason mismatch*, *Competitor mentioned* · *New prices*, *Mentioned in period*, *Awaiting approval*, *Companies*.
- **Highlights** and **Suggested steps** (**To discuss** in the 1:1 brief) — rule-based; with the AI summary on, an *Executive summary* at the top.
- **Charts** — by owner, warning, stage, reason, push length, month.
- **Tables** — at-risk deals (with reason and suggested step), recovered deals, repeat slippers, lost deals, competitor prices; in the 1:1 brief a **brief card** per rep (period summary, standing against target, questions).
- **Play**, **Video**, **Print**, **Run again**, **Agent**.

**Number = list:** clicking an indicator, a chart slice or a number on a brief card opens the records behind it. The report keeps the records as they were when it was produced; the list shows them as they are today.

The comparison with the previous report (newly at risk / recovered, losses against the previous period) starts from the second report.

## Notification {#notify}

When a report is ready, an **Agent reports** notification arrives — bell, email and phone ([Notifications](/en/guides/notifications)); the title is *Agent report · <agent name>*. The email carries the summary, highlights and suggested steps, and the link opens the report page. The person who runs a report by hand isn't notified.

A report can be seen by the person who set up the agent, its recipients and admins.

## Competitor prices {#prices}

Prices found by the **Competitor price log** agent are collected on **Agents › Competitor prices** (`/ajanlar/rakip-fiyatlari`): competitor, price (with currency and unit), product, company, a quote from the conversation and the source (call recording / visit note).

- **Approve** — the price is written to the competitor details on the company card (the competitor is added if missing; price note *€3.80/kg · 20.10.2026 (from conversation)*). **Reject** — not saved. Approvals can be made by admins, the person who captured the price in the conversation and the company owner; admins can **Approve visible** in bulk.
- The same price from the same conversation isn't added twice.
- At the top, **competitors' latest prices**: latest price, the one before and the change (%).
- With *Save prices without approval* on, prices are written without waiting for approval.

Prices come from the price signals of [conversation intelligence](/en/guides/conversation-intelligence#signals): a sentence with a price is linked to the competitor named in the same or the previous sentence.

## AI summary {#ai}

With AI on, the report's figures (indicators, highlights, table rows) are sent to the model, which is asked for a short executive summary and suggested steps. **Every number the model writes is looked up in the report's figures**; a sentence or step with a number that isn't in the report is dropped. If the model doesn't respond, the report comes with its rule-based summary (noted on the report). People's phone numbers and email addresses aren't sent.

## API {#api}

- [`GET /agents/runs`](/en/rest-api/agents/runs) — the latest reports you can see (`limit`, max 100).
- [`GET /agents/runs/{id}`](/en/rest-api/agents/run) — a report: indicators, highlights, suggested steps, AI summary, tables, briefs.
- `caps.agents` in the `/me` response.

## Limits {#limits}

- Agents can't be extended with new templates (the five ready-made agents); one scope per agent.
- The scheduler checks every 10 minutes; up to five agents run at a time, the rest run on the next round.
- Target and quota figures in the 1:1 brief come when the Forecast module is on.