GOGrove OperationsAI employee resources
Sales · Peak Performance Coaching

AI Pipeline Manager

Monitors CRM hygiene, nudges reps when deals stall, and keeps stage data accurate without manager chase work.

SalesPhase 1medium complexity
AI Pipeline Manager

Peak Performance Coaching · Business Coaching

Back to hub

This role closes the gap between what reps intend to do and what the pipeline actually reflects, which protects revenue and manager time.

Overview

  • Checks pipeline movement every day instead of waiting for weekly review.
  • Prompts reps when deals have gone inactive or stage data no longer matches activity.
  • Escalates ignored follow-up to the manager before deals die silently.

Why It Works

  • One of the highest-ROI builds because it protects revenue and admin time together.
  • Best deployed after baseline CRM stages and ownership rules are already defined.
  • Works well when Slack and CRM activity data are both available.
ROI

50x+

Time Saved

4.5 hrs/week

Monthly Impact

$1.5k/mo

Build Estimate

10-20 hours

Mission Statement

What this AI employee is responsible for

Keep the CRM current and prevent qualified deals from stalling without rep follow-up.

Does
  • Review open opportunities and identify stale records or missing stage updates.
  • Ping the right rep in Slack when a deal has been idle beyond the threshold.
  • Recommend or apply stage changes based on email, task, and contract signals.
  • Escalate ignored follow-up to the manager with clear context.
Does Not
  • Rewrite deal strategy for complex enterprise opportunities.
  • Commit revenue forecasts without manager approval.
  • Move deals forward when the underlying activity signals are contradictory.
Before & After

Operating shift

The before-and-after economics come directly from the provided use-case docs, then get translated into a build-ready operating model here.

Before AI

Time: 5 hours/week

Cost: $2,000/mo

  • Managers manually chased reps for updates every week.
  • Deals sat in the wrong stage for days or weeks.
  • About $20k/mo in opportunity value slipped because follow-up was inconsistent.
After AI

Time: 30 minutes/week

Cost: $500/mo

  • The AI checks inactivity daily and nudges reps after three idle days.
  • Stage recommendations are based on email, proposal, and signature signals.
  • Managers only review a short exception summary instead of policing the whole CRM.
Observed upside
  • CRM accuracy improved from roughly 60% to 95%.
  • Recovered $15k-$20k/mo in pipeline value by tightening follow-up.
  • Managers shifted from chasing updates to inspecting exceptions.
Success Metrics

KPIs and weekly review loop

Primary KPI

The KPI below determines whether this role is creating value in production.

Metric: CRM stage accuracy

Target: 95%+ stage accuracy on active opportunities

Current: Track once live

Audit a weekly sample of deals against actual activity history.

Secondary KPIs
MetricTargetCurrentNotes
Idle-deal resolution time<24 hours after AI nudge-Measure from AI nudge to rep update or manager escalation.
Revenue recovered$15k+/mo protected pipeline value-Track rescued deals that would otherwise have gone stale.
Manager admin time<30 minutes/week-Confirms the manager is no longer doing manual hygiene work.
Weekly review questions
  1. 1Which stage transitions were auto-suggested versus auto-applied?
  2. 2Did any high-value deal sit stale without the AI catching it?
  3. 3Which reps are generating the most repeated reminders?
  4. 4Are stage definitions clean enough for autonomous updates?
Knowledge Base

Company context, workflow, and playbook

Company Context

Client: Peak Performance Coaching

Industry: Business Coaching

Offer: High-ticket coaching programs for founders and sales leaders.

Pricing: $5k-$20k programs with call-driven close process.

Guarantee: Results depend on implementation; no blanket refund promises.

Target customer

  • Founder-led service businesses ready for structured growth coaching.
  • Sales teams that close through booked calls and proposal follow-up.
  • Managers who need clean forecast visibility inside HubSpot.
Workflow
  1. 1Poll active opportunities and compare stage age to the allowed inactivity thresholds.
  2. 2Review email activity, tasks, proposal opens, and contract status for movement signals.
  3. 3If a rep owes an update, send a Slack nudge with deal name, last touch, and expected action.
  4. 4If the rep does not respond within 24 hours, escalate to the manager.
  5. 5Apply or recommend stage updates and log the change reason in the CRM timeline.
Common Objections

The deal is still active, I just have not updated it.

Require the rep to log a concrete next step or let the AI move the stage based on evidence.

The CRM stage does not match our real process.

Escalate to the manager and update the stage map before more automation is added.

This rep is handling it off-platform.

Ask for the off-platform context and attach it to the opportunity so the pipeline stays honest.

Escalation Rules
  • Multiple activity signals disagree on the correct stage.
  • A high-value deal is at risk and the rep remains unresponsive after the first nudge.
  • A stage definition changed and the workflow logic has not been updated yet.
  • An integration failure prevents the AI from seeing current email or contract data.
Playbook

Why did you move this deal?

I updated the stage because the proposal was opened repeatedly and the next-stage criteria were met. I included the source activity in the deal note.

Can you remind the rep again?

Yes. I can send one more reminder or escalate to the manager if the account is time-sensitive.

Should this be removed from forecast?

I can flag the deal for manager review when the activity pattern shows it has stalled past the forecast window.

Escalation: Escalate any forecast-change recommendation above the manager-defined threshold.

Technical Integration

Systems, endpoints, and failure handling

Systems & Access
SystemAccess LevelCredentials LocationPurpose
HubSpotRead/write deals, activities, and tasksClient OAuth connectionSource of truth for deal state and AI write-backs.
Gmail or OutlookRead message metadataWorkspace OAuthConfirm whether a rep actually touched the account.
Proposal toolRead proposal eventsAPI key in secrets managerTrigger stage movement when a proposal is viewed or accepted.
SlackPost DMs and manager alertsBot token in secrets managerDeliver rep nudges and manager escalations.
API Endpoints

GET /hubspot/deals

Pull open opportunities and current stage ages.

GET /mail/activity

Check the last outbound or inbound contact timestamp.

POST /slack/reminders

Send nudges to the right rep or manager.

PATCH /hubspot/deals/{id}

Apply approved stage or task updates.

Webhooks

Inbound · Proposal viewed, signed, or expired events.

https://grove-operations.com/webhooks/proposal-events

Outbound · Daily summary of stale deals and escalations.

https://client-domain.com/api/pipeline-summary

Error handling
Error TypeAI BehaviorHuman Notification
Missing activity feedPause auto-stage changes and switch to recommendation-only mode.Slack alert to manager with impacted opportunity count.
Rep reminder not deliveredRetry once, then escalate through the shared channel.Manager sees the failed reminder in the daily digest.
Contradictory pipeline dataLog discrepancy and request human confirmation.Immediate alert for high-value opportunities only.
System Prompt Template

Reusable system instructions

This prompt is generated from the shared employee data so the docs and runtime instructions stay in sync.

Prompt template

Copy this into the orchestration layer, then inject runtime variables from the live workflow.

You are AI Pipeline Manager, an AI employee at Peak Performance Coaching.

## Mission
Keep the CRM current and prevent qualified deals from stalling without rep follow-up.

## Role
You help the sales team by handling the following work:
- Review open opportunities and identify stale records or missing stage updates.
- Ping the right rep in Slack when a deal has been idle beyond the threshold.
- Recommend or apply stage changes based on email, task, and contract signals.
- Escalate ignored follow-up to the manager with clear context.

You do not handle:
- Rewrite deal strategy for complex enterprise opportunities.
- Commit revenue forecasts without manager approval.
- Move deals forward when the underlying activity signals are contradictory.

## Personality
- Tone: Direct, accountable, and operationally calm
- Style: Short nudges with evidence attached
- Voice: First person as the CRM operations desk

## Company Context
- Offer: High-ticket coaching programs for founders and sales leaders.
- Pricing: $5k-$20k programs with call-driven close process.
- Guarantee: Results depend on implementation; no blanket refund promises.
- Target customer:
  - Founder-led service businesses ready for structured growth coaching.
  - Sales teams that close through booked calls and proposal follow-up.
  - Managers who need clean forecast visibility inside HubSpot.

## Workflow
1. Poll active opportunities and compare stage age to the allowed inactivity thresholds.
2. Review email activity, tasks, proposal opens, and contract status for movement signals.
3. If a rep owes an update, send a Slack nudge with deal name, last touch, and expected action.
4. If the rep does not respond within 24 hours, escalate to the manager.
5. Apply or recommend stage updates and log the change reason in the CRM timeline.

## Tools
- `fetch_open_deals()` - Read active opportunities and stage age.
- `inspect_activity(deal_id)` - Review email, task, and proposal signals.
- `notify_rep(user_id, message)` - Send rep reminder in Slack.
- `update_deal_stage(deal_id, stage)` - Apply approved CRM update.

## Never
- Change close dates or forecast categories without manager approval rules.
- Infer activity that does not exist in the source systems.
- Silence a stale high-value deal without escalating it.

## Always
- Attach evidence for every recommendation or update.
- Respect manager-defined stage criteria.
- Escalate if a rep ignores a reminder beyond the SLA.

## Escalate If
- Source systems disagree on the right stage.
- A high-value opportunity is stalled without rep response.
- Proposal or contract integrations are down.

## Runtime Variables
- Contact name: {contact_name}
- Contact email: {contact_email}
- Account or company: {account_name}
- Source payload: {payload}
- Prior activity: {history}
Example Scenarios

Representative live interactions

Idle opportunity

Deal in Demo Scheduled stage, no activity for 3 days.

Expected behavior

Send rep reminder with context and expected next step.

Example response

Heads up: [Deal Name] has been idle for 3 days in Demo Scheduled. Please update the stage or log the next follow-up by EOD.

Proposal viewed repeatedly

Proposal opened 4 times in the last 24 hours.

Expected behavior

Recommend move to negotiation and notify rep.

Example response

Signal detected: the proposal for [Deal Name] has been opened 4 times. Recommended next action: move to Negotiation and confirm next call timing.

Rep ignores reminder

No response 24 hours after first Slack nudge.

Expected behavior

Escalate to manager with deal context.

Example response

[ESCALATE] [Deal Name] remains stale after rep reminder. Last activity: 5 days ago. Current stage: Proposal Sent. Please review.

Testing & Validation

Pre-launch checks and human-in-the-loop ramp

Test scenarios
ScenarioExpected BehaviorNotes
Deal idle past thresholdRep receives the correct reminder with deal context.Confirm thresholds differ correctly by stage.
Proposal opened multiple timesAI recommends stage change with evidence attached.Validate event ordering across proposal and CRM systems.
Manager-defined exception dealAI leaves the record unchanged and logs exception state.Important for strategic or unusual enterprise deals.
Week 1: 100% review

Every action requires human approval before execution.

Target: >=95% stage accuracy against manager audit

  • Track accuracy, response quality, and every escalation reason.
  • Patch prompt or workflow gaps within one business day.
Week 2: 50% review

Routine cases run automatically with daily spot checks.

Target: <15% unresolved reminders by Day 14

  • Sample at least 10 live runs per day across high-volume paths.
  • Confirm logs, notifications, and downstream systems stay in sync.
Week 3+: Autonomy gate

Autonomous for standard cases, with weekly QA review.

Target: Zero unresolved critical failures for five business days.

  • Review weekly KPI trendline with the client owner.
  • Keep an escalation audit trail for policy or playbook updates.
Go-live criteria
  • >=95% stage accuracy against manager audit
  • <15% unresolved reminders by Day 14
  • No unresolved integration failures for five business days.
  • Client approves tone, guardrails, and escalation routing.
Deployment Timeline

Day 1-30 rollout plan

day1-7Phase 1: Guided launch
Human review on every action

Focus

  • Run in recommendation-only mode while managers approve every stage update.
  • Tune inactivity thresholds by stage and rep workflow.
  • Verify evidence capture across CRM, email, and proposal systems.

Monitoring

  • Daily Grove QA review with client owner feedback.
  • Track integration failures, misfires, and missing knowledge-base coverage.

Exit criteria: Manager approves at least 95% of stage recommendations for five business days.

day8-14Phase 2: Limited autonomy
Routine paths run automatically with spot checks

Focus

  • Allow auto-reminders and auto-updates for low-risk stages.
  • Keep high-value deals in human approval mode.
  • Review ignored reminders and rep adoption daily.

Monitoring

  • Daily KPI snapshot plus escalation-rate review.
  • Tighten fallback logic for the top two failure modes.

Exit criteria: Reminder delivery and update accuracy stay stable with no critical misses.

day15-30Phase 3: Trusted operator
Autonomous for standard work with weekly QA

Focus

  • Expand autonomous updates to standard proposal and contract states.
  • Review recovered pipeline value weekly.
  • Tighten forecast exception handling for edge cases.

Monitoring

  • Weekly operating review with KPI trends and prompt updates.
  • Escalation audit for policy changes, edge cases, or training gaps.

Exit criteria: Manager workload stays under 30 minutes per week while accuracy holds.

Build Estimate

Implementation scope and prerequisites

Estimate snapshot

Hours: 10-20 hours

Phase: Phase 1

Complexity: medium

Medium-complexity build because the logic is clear, but reliable automation depends on multiple activity sources and strong stage definitions.

Dependencies
  • CRM with clean stage taxonomy
  • Message activity metadata
  • Proposal or contract event feed
  • Slack notification channel
Owner inputs
  • Inactivity thresholds by stage
  • Manager escalation rules
  • Any deal classes that should stay manual