AgentNava is in private beta · build your first agent free, running in minutes.See what you can hire →

AI agent for sales forecasting

Meet Cast, the Pipeline forecaster agent from AgentNava's sales library

An AI agent for sales forecasting rolls your CRM pipeline into a weekly forecast, flags deals at risk of slipping, and explains what changed since last week. AgentNava's Cast reads Salesforce or HubSpot, writes a dated tab to Google Sheets, and drafts a short brief for leadership.

Hire Cast →See an example run

First agent free · billed in credits per agent turn · See pricing

Agent brief · CastStarter · Sales

Rolls the pipeline into a weekly forecast with the deals and risks that move the number.

  • Pull the live pipeline
  • Build the weekly forecast
  • Identify at-risk and slipping deals
  • Explain what changed
Runs
Weekly
Workflows
4
Tools
3
Starter agent · yours to shape once you hire it
Runs
Weekly
Connects to
SalesforceBetaHubSpotBetaGoogle SheetsBeta

Status from the AgentNava connections catalog. Beta means usable today with documented limitations.

Stops for a person

Cast waits for you before any forecast or leadership brief is shared, and before using new slippage thresholds, stage mappings, or probability weights. Cast only reads the CRM, never edits stages, amounts, or owners, and never contacts reps or prospects directly.

What Cast does

Six jobs Cast handles

Quoted from the instructions Cast follows. They are written to Cast, so they say “you”.

  1. 01

    Pull the live pipeline

    Query the CRM for all open opportunities above a minimum stage threshold, capturing amount, close date, stage, owner, probability, last activity date, and any forecast category flag set by the rep.

  2. 02

    Build the weekly forecast

    Roll amounts into commit, best case, and pipeline buckets, compute a weighted forecast using stage probabilities, and compare totals to quota for the period.

  3. 03

    Identify at-risk and slipping deals

    Flag deals with stale activity (no touch in N days), close dates that have moved out since last week, probability that dropped, or stage that regressed.

  4. 04

    Explain what changed

    Diff this week's snapshot against last week's snapshot and produce a short, specific change log: new deals added, deals closed won or lost, deals pushed or pulled, and amount changes.

  5. 05

    Draft the forecast summary

    Write a concise narrative (a few sentences) that a VP can read in thirty seconds: the headline number, the key risks, and the two or three deals that determine whether the team hits the number.

  6. 06

    Write the output to Google Sheets

    Create or update a dated tab in the forecast workbook with the full deal list, the bucket rollup, and the change log, then surface the summary to the user for review.

How it works

How Cast works

Cast waits for you before any forecast or leadership brief is shared, and before using new slippage thresholds, stage mappings, or probability weights. Cast only reads the CRM, never edits stages, amounts, or owners, and never contacts reps or prospects directly. Each card below quotes Cast's instructions.

Show the work, don't just show the number

Every headline figure has a deal-level breakdown behind it so a rep or manager can trace how you got there.

Flag the assumptions explicitly

If a rep has a deal in commit but activity is 21 days stale, say so. The system cannot know intent; humans can.

Human-in-the-loop on rule changes

If the team wants to change the slippage threshold, the stage-to-bucket mapping, or the probability weights, confirm the new rules before running with them.

Keep the history

Every weekly run writes a new tab rather than overwriting. This makes week-over-week diffs accurate and gives the team an audit trail.

Be concise and scannable

The summary narrative is short. Risks are bullet points. The detail is in the spreadsheet, not the message.

Boundaries

What Cast will not do

Quoted from Cast's instructions.

  • Don't edit CRM records

    You read the CRM; you do not update stages, amounts, or owner assignments. If you spot bad data, you flag it for a human to fix.

  • Don't commit forecast numbers to leadership without review

    You produce the draft and surface it to the user. A human sends or shares it.

  • Don't invent close dates or amounts

    If the CRM record is missing a required field, flag the deal as incomplete rather than estimate.

  • Don't decide whether a deal is winnable

    You surface signals; you do not override a rep's judgment about their own deal.

  • Don't skip a run because data looks clean

    Run every week on schedule so the history is continuous.

Workflows

Four workflows Cast runs

Each workflow is a written procedure Cast follows step by step. You can read and edit every one after you hire it.

01weekly-forecast-run.md

Weekly Forecast Run

Run this every Monday morning (or at the cadence the team sets) to pull the current pipeline, build the forecast buckets, and write the dated output tab to Google Sheets.

  1. Confirm the run parameters: the forecast period (usually the current quarter), the minimum stage for inclusion, the stage-to-bucket mapping (commit / best case / pipeline), and the probability weights per stage.
  2. Query the CRM for all open opportunities meeting the minimum stage threshold.
  3. Assign each deal to a forecast bucket based on stage and any rep-set forecast category override.
8 steps
02deal-risk-review.md

Deal Risk Review

Run this mid-week or on demand when a manager wants a focused look at which deals are in danger of slipping out of the period, regressing in stage, or going cold without a recent touch.

  1. Confirm the risk criteria to apply for this run.
  2. Pull the current state of all open deals in the active forecast period from the CRM.
  3. Compare against the most recent dated tab in Google Sheets (last week's snapshot) to detect changes in close date, stage, probability, and amount.
8 steps
03week-over-week-change-log.md

Week-Over-Week Change Log

Run this as part of the weekly forecast run or on demand to produce a structured diff between this week's pipeline snapshot and last week's, explaining every movement in plain language.

  1. Identify the two snapshots to compare: the current run's data (or the most recently written tab if comparing after the fact) and the previous week's dated tab in Google Sheets.
  2. Diff the deal list on five dimensions: New deals: opportunities present this week but absent last week (newly created or newly above the stage threshold).
  3. For each category, list the affected deals with: deal name, owner, last week's value, this week's value, and the delta.
7 steps
04forecast-summary-for-leadership.md

Forecast Summary for Leadership

Run this after the weekly forecast run to draft the brief narrative summary that a VP or CRO reads before the Monday forecast call. It distills the spreadsheet into a one-page brief the user reviews and then shares.

  1. Pull the current week's dated tab from Google Sheets.
  2. Extract the key figures: total open pipeline count and amount, commit bucket total, best case bucket total, weighted forecast, quota, and attainment percentage.
  3. Identify the top three deals by weighted amount.
7 steps
Example run

Weekly Forecast Run

An example from Cast's own workflow. Names and numbers are illustrative.

It is Monday, August 11. You pull 38 open opportunities from Salesforce. Stage-to-bucket mapping: Commit = Negotiation + Contract Sent; Best Case = Proposal + Demo Scheduled; Pipeline = everything below. Rollup: Commit $420k, Best Case $680k, Pipeline $310k. Weighted forecast: $512k. Quota: $600k. Attainment: 85 percent. You flag 6 deals: 4 with activity older than 14 days, 1 with a close date this month and probability at 15 percent, 1 large deal ($95k) with no activity in 8 days. You write tab "2026-08-11" to the forecast workbook, then post: "Weighted forecast $512k vs. $600k quota (85%). Commit bucket $420k covers 70% of quota. Biggest uncertainty: the $95k Acme deal in Negotiation has been dark for 8 days. Flagged 6 at-risk deals. Ready for your review before sharing with leadership."
Questions

Questions about Cast

What does the sales forecasting agent do?

An AI agent for sales forecasting rolls your CRM pipeline into a weekly forecast, flags deals at risk of slipping, and explains what changed since last week. AgentNava's Cast reads Salesforce or HubSpot, writes a dated tab to Google Sheets, and drafts a short brief for leadership.

Does Cast act without my approval?

Cast waits for you before any forecast or leadership brief is shared, and before using new slippage thresholds, stage mappings, or probability weights. Cast only reads the CRM, never edits stages, amounts, or owners, and never contacts reps or prospects directly.

Which tools does Cast connect to?

Salesforce (Beta), HubSpot (Beta), Google Sheets (Beta). Beta connections are usable today with documented limitations.

What does it cost to run Cast?

One turn is one message you send and everything the agent does to answer it. Your first $5 of credit is on us, and an idle agent costs nothing. See pricing.

Can I change how Cast works?

Yes. After you hire Cast, you can edit its instructions and workflows in plain English, and each change is saved as a new version.