The problem
Watch any B2B GTM weekly account review. It's a status-update theater. AE says "spoke with the champion, they're aligned, waiting on legal." Sales manager says "ok, push." 30 minutes, no actual decisions.
The meeting doesn't fail because the team is bad. It fails because: - AE has 12 deals; manager has 80 across the pod. Nobody has live context on all of them. - Status updates are AE narratives, not pipeline analysis. - Decisions get deferred because data isn't on the table.
The fix: 4-minute AI prep produces a 1-page deal review packet
Before the weekly sync, an AI script runs across every active opportunity in the pod and produces a single 1-page packet with three columns:
| Deal | Status (AI-derived) | Decision needed |
The "status (AI-derived)" column is not what the AE typed in Salesforce. It's the model's analysis of: most recent Gong call, last 14 days of email, MAP (mutual action plan) progress, competitive mentions, time-in-stage anomaly.
The "decision needed" column is the killer feature. AI flags: - Deals where AE narrative diverges from data (AE says "verbal yes", emails show 4-day silence) - Deals where the buyer center has shifted (champion changed roles, new exec entered, decision-maker quiet) - Deals where competitive risk surfaced (Gong picks up vendor names in last call) - Deals where MAP is stalling (commits made, dates slipping)
What the meeting becomes
Instead of going around the table, the manager opens the packet. Says: "I see 3 deals where AI flags status divergence. Let's start there."
Each flagged deal gets 5 minutes. Decision made. Action item assigned. Move on.
The other deals — no flags — get a 30-second AE confirm or correction. Done.
A 30-minute meeting becomes a 20-minute decision factory.
The Claude Code skill
name: weekly-pod-review
trigger: Cron job — every Sunday 7pm
Inputs (per opportunity): - Salesforce: stage, amount, close date, MAP fields - Gong: last call transcript + AI summary - Email: thread analysis last 14 days - Clari (if you have it): forecast confidence
Output: Slack message in the pod channel + Notion page.
The system prompt
You are a sales operations analyst. For each open opportunity in the pod, produce one row with:
- Deal name (account / opportunity ID)
- Status (AI-derived) — one short sentence based on actual data, NOT the AE's notes
- Decision needed — one specific action item OR "none — proceeding normally"Flag a deal as "decision needed" if any of: - AE narrative diverges from data (AE optimism + buyer silence) - Buyer center shift (new contact, departed contact, role change) - Competitive mention surfaced in last 14 days - MAP commitment slipped > 5 days - Time-in-stage > 1.5x team median - Last activity > 10 days ago for stage 4+
Output as a markdown table, sorted: decision-needed deals first. Max 25 rows. If pod has more than 25 active deals, surface the 25 most-at-risk. ```
Specific behaviors to enforce
- **No AE writing**: AE doesn't update the packet. The packet is derived from system data. If AE wants to add commentary, do it in the meeting verbally.
- **Manager prep is mandatory**: 5 minutes before the meeting, manager reviews packet, picks 3 deals to focus on.
- **Decisions logged**: every flagged deal gets an action item in Salesforce next-step field. Reviewed next week.
Pitfalls
- **Model surfaces wrong divergence**: this happens when AE narrative is right and the email data lags. Solution: AE can flag "false positive" in next-step field; this trains the prompt over time.
- **Pod resists the change**: replacing narrative theater with data-driven review feels surveillance-y. Frame as "we're freeing 10 minutes for actual coaching, not status updates."
- **Sales manager doesn't trust the AI**: they're right to be skeptical initially. Run packet alongside traditional sync for 4 weeks, compare action items + outcomes.
Tools
- **Gong / Chorus**: call data
- **Salesforce / HubSpot**: opp + MAP data
- **Clari**: forecast + confidence data
- **Anthropic Claude**: long-context analysis (often 50+ deals worth of data in one prompt)
- **Slack**: distribution
- **Notion** (optional): pinned page for the week
What changes
- Meetings drop 10-15 minutes
- Manager coaches more, status-updates less
- Surfaced risks get caught 2-3 weeks earlier
- AE morale improves (less performative)
Why this is a handoff article
This isn't a single role's problem. It's the seam between AE (deal owner) and RevOps / sales manager (deal portfolio owner). AI bridges the visibility gap that costs deals.