Build This: An AI Morning Briefing That Tells Your CSMs Where to Focus

Here's how most CSMs start their day: open the CS platform, scroll through their portfolio, scan dashboards, check email for customer responses, peek at the support tool for new tickets, and try to mentally prioritize 60-100 accounts based on whatever they happen to notice first.

This process takes 30-45 minutes. It's inconsistent. It's biased toward squeaky wheels. And it misses the quiet accounts that are silently drifting toward churn.

You can replace the entire ritual with a single automated briefing delivered to each CSM before they open their laptop. Here's exactly how to build it.

What the Briefing Looks Like

Every morning at 8am, each CSM receives a Slack message (or email) that says:

Good morning, Sarah. Here are your top 5 accounts for today.

1. Acme Corp — URGENT Health score dropped from 72 to 58 in the last 7 days. Primary driver: usage declined 34% week-over-week. Renewal in 67 days. No QBR scheduled. Recommended: Schedule a check-in this week. Lead with their Q4 goals and probe for adoption blockers.

2. Beacon Analytics — WATCH Support ticket volume spiked: 4 tickets in 5 days (vs. their average of 1/month). Sentiment on last two tickets: Negative. Champion (Jamie Torres) hasn't responded to your last email (sent 8 days ago). Recommended: Call Jamie directly. Don't email. Reference the support issues and ask what's going on.

3. CloudNine Systems — EXPANSION Usage at 94% of licensed capacity for 3 consecutive weeks. Two new departments onboarded in the last 30 days. Renewal in 120 days. Recommended: Expansion conversation. They're outgrowing their current plan. Propose an upgrade before renewal.

4. DataForge Inc — RISK Champion (Alex Kim) updated LinkedIn: new role at a different company. Last internal meeting was 42 days ago. No other stakeholder with a Relevance Score above 40. Recommended: Identify a new champion. Ask your contact for an introduction to their successor before you lose the thread.

5. Elevate HR — FOLLOW-UP QBR was 3 days ago. Action items logged but no follow-up sent. Sentiment from QBR: Positive. Recommended: Send the follow-up brief today while momentum is fresh.

That's 5 paragraphs. Takes 2 minutes to read. The CSM sits down knowing exactly where to focus and why. No dashboard scanning. No guesswork.

The Architecture

Three components: data aggregation, AI processing, and delivery.

Component 1: Data Aggregation

You need a scheduled job that pulls fresh data from all relevant sources every morning. In a Planhat-centric stack:

From Planhat API:

• Company health scores (current + 7-day change)
• License data (renewal dates, ARR, product)
• End User data (Relevance Scores, last activity date)
• Recent Conversations (sentiment, category)
• Tasks and Notes (last CSM activity)
• Calculated Metrics (usage velocity, deviation from normal)

From Salesforce API:

• Opportunity pipeline (expansion deals, renewal forecast stage)
• Activity data (meetings, emails logged against the account)
• Contact changes (new contacts added, contacts removed)

From Support Tool API (Zendesk/Intercom):

• Open tickets (count, severity, age)
• Recent closed tickets (resolution, satisfaction rating)
• Ticket volume trend (this week vs. average)

Pull this data into a structured JSON object per account. Each account gets a snapshot that looks something like:

{
  "company": "Acme Corp",
  "health_score": 58,
  "health_score_7d_change": -14,
  "arr": 45000,
  "renewal_days": 67,
  "usage_change_30d": -0.34,
  "last_csm_activity_days": 12,
  "open_tickets": 1,
  "ticket_volume_trend": "normal",
  "champion_name": "Lisa Park",
  "champion_relevance": 82,
  "champion_last_activity_days": 5,
  "sentiment_30d": "neutral",
  "qbr_scheduled": false,
  "expansion_signals": false
}

Component 2: AI Processing

Feed the aggregated data for all accounts in a CSM's book to Claude (via API) or through Planhat's MCP Server. The prompt structure:

You are a CS operations analyst. Below is the current data for all
accounts in [CSM Name]'s portfolio.
Rank the top 5 accounts that need attention today, in order of urgency.
For each account:
1. State the account name and urgency level (URGENT/WATCH/EXPANSION/
   FOLLOW-UP/RISK)
2. Explain what changed and why it matters, in 2-3 sentences
3. Recommend a specific next action
Prioritization criteria:
- Health score declines > 10 points in 7 days = highest urgency
- Renewals within 90 days with risk factors = high urgency
- Champion departure or disengagement = high urgency
- Support sentiment degradation = medium urgency
- Expansion signals = medium (positive action)
- Overdue follow-ups = low urgency
Here is the data:
[account data array]

Claude returns a structured briefing for that CSM. Run this for each CSM in parallel.

Component 3: Delivery

The output goes to wherever the CSM already works:

• Slack: Post to a DM or a dedicated #cs-briefings channel
• Email: Simple formatted email delivered at 8am
• Planhat: Log the briefing as a Note on a CSM-specific Project or Dashboard

Use Planhat's AI Workflows to orchestrate the whole chain: scheduled trigger → data pull → AI processing → output delivery. Or build it as a standalone script (Node.js or Python) that runs on a daily cron.

What Makes This Work

It's proactive, not reactive. The CSM doesn't have to go looking for problems. The problems come to them, ranked and contextualized.

It catches silent risk. The CSM who manually scans dashboards gravitates toward familiar accounts and recent interactions. The automation surfaces the account that hasn't been touched in 42 days where the champion just left. That's the one that was about to churn while nobody was looking.

It's personalized per CSM. Each CSM gets their own briefing based on their specific book. Not a team-wide report that nobody reads.

It compounds over time. The AI processing can learn which recommendations the CSMs act on and which they ignore. Over time, refine the prioritization criteria based on actual behavior and outcomes.

Build Time

If your data foundations are solid (see my post on AI-ready data architecture), this takes about 2 weeks to build and test:

• Week 1: Data aggregation pipeline, API connections, structured data format
• Week 2: Prompt engineering, output formatting, delivery integration, CSM feedback loop

If your data has gaps (usage not flowing in, contacts not mapped, no time-series history), fix those first. The briefing is only as good as the data behind it.

The ROI

5 CSMs x 35 minutes saved per morning x 5 days a week = 14.5 hours recaptured weekly. That's ~750 hours per year of CSM capacity redirected from dashboard scanning to customer-facing work.

But the bigger number is the saves. One account caught early that would have churned silently is worth $30K, $50K, $100K depending on your ACV. The morning briefing pays for itself the first time it surfaces a risk you would have missed.