# Customer Success

> Onboarding automation, health monitoring, renewal and expansion, and async check-ins.

Customer success is where the money you already won either compounds or leaks. Acquisition is loud. Retention is silent — until it isn't, and the account churns the week before renewal because nobody noticed it went quiet. Agents fix this. They are tireless, never forget a client, and surface signals the human eye misses. The whole game in CS is seeing the account clearly and acting on what you see before the client has to ask. An agent operating across your email, calendar, project tools, and product data does that on autopilot. A few principles before the tips: Quiet is a signal, not a status. If a client hasn't talked to you in three weeks, that's information. Treat it like a flashing yellow. Done-for-you beats done-with-you. Clients pay for outcomes, not for meetings. Async reports they can watch at 11pm beat a Zoom they have to schedule. The agent is the system of record. If it lives in your head, it doesn't exist. Get every commitment, deliverable, and follow-up into one place the agent can read.

## 1. Onboarding Automation

The first 14 days set the entire relationship. A client who hits a small win in week one renews 3x more often than one who doesn't. Onboarding is also the most repetitive thing CS people do — same welcome email, same kickoff checklist, same "here's your Loom intro" twelve times a month. Pure agent territory.

### Tip 1.1 — Auto-fire welcome sequence on deal close

**What it does:** The moment a deal moves to "won" in your CRM, your agent kicks off a 7touch welcome sequence: welcome email, calendar booking link for kickoff, environment setup checklist, a personalized Loom intro using your face/voice, two value emails over the first week, and a check-in at day 14. Why it wins: Clients form their loyalty in the first two weeks. Most agencies/SaaS shops botch this because the AE hands off to CS, CS isn't ready, and the client sits in silence for 5 days. Silence in week one is the single biggest churn predictor in B2B services. Tools: Your CRM (Attio, HubSpot, Notion DB), Gmail or your email API, Cal.com or Calendly for the kickoff booking, Loom for the intro video, your voice email skill.

**How to wire it:** 1. Define the stage trigger in your CRM. "Won" → fires a webhook to your agent. 2. Map the 7 touches to specific days: T+0 welcome, T+1 setup checklist, T+2 kickoff booking nudge, T+5 first value email, T+9 second value email, T+14 status check-in. 3. For each touch, the agent loads context (deal notes, kickoff call transcript if it exists, deliverables agreed) and drafts the message in your voice using your email voice skill. 4. Drafts land in your outbox. You approve and send — or for trusted clients, auto-send with a "reply STOP to pause" line. 5. The agent tracks opens/replies and pauses the sequence the moment the client engages with anything. Example prompt to your agent: Listen for deal_stage = won webhook from Attio. When fired, kick off the new-client onboarding sequence: load the deal notes, the kickoff call transcript if one exists, and the agreed deliverables. Draft 7 touches at T+0, +1, +2, +5, +9, +14, +21 using my email-voice skill. T+0 is a warm welcome with a Cal.com link for kickoff. T+1 is the env setup checklist (use client-onboarding skill). T+2 nudges them if they haven't booked. T+5 and T+9 are value emails — link to a relevant case study, no ask. T+14 and T+21 are status check-ins. Stage every draft in my client-onboarding/<clientslug>/ folder and ping me each morning with what's queued. Pause the sequence the second the client replies. Watch out for: Don't let the agent send case studies that aren't relevant. Tag your case studies by ICP and have the agent pick. The kickoff Loom should feel personal, not templated. Re-record it for any deal above $20k MRR. Email deliverability: if you're sending all 7 from a fresh domain, you'll hit spam. Warm the domain first. Skill file: client-onboarding, email-followups, call-confirmation-emails

### Tip 1.2 — Environment & access setup checklist (per-client)

**What it does:** The agent builds a per-client setup checklist based on what you actually need from them (API keys, Slack invite, calendar share, doc access, etc.) and runs it as a structured task list. It chases missing items, drafts the reminder, and only marks the client onboarded when every box is ticked.

**Why it wins:** Half-onboarded clients haunt you for weeks. The CS person forgets that the Calendly integration was never finished, and three weeks in the client says "I thought this was supposed to do X" — and they're right. A structured checklist with the agent chasing the gaps means nothing slips. Tools: A checklist template in Markdown or Notion, your email and Slack channels, your CRM for status, optionally an agent-browser instance if you need to walk the client through a UI on a Loom. How to wire it: 1. Define a master checklist template for your offering. For each item: what's needed, who provides it, how it's verified. 2. The agent instantiates the checklist for each new client at kickoff, fills in the easy items (your side), and emails the client a clean list of their side. 3. Daily cron: agent checks each open item. Verified? Tick. Pending more than 48h? Draft a friendly nudge in your voice. 4. When everything green, the agent fires a "you're fully set up" email and updates the deal stage in the CRM. Example prompt to your agent: For each new client, instantiate the master onboarding checklist at templates/onboarding-checklist.md . Fill in my-side items (Slack channel created, Drive folder shared, kickoff booked). Email the client a clean list of their-side items with verification steps. Save the checklist to clients/<slug>/onboarding.md . Every day at 9am, check each open item against verification (API key works, Slack invite accepted, doc shared). For anything pending 48+ hours, draft a nudge in my voice. When all items green, send the "you're fully set up" email and move the deal to onboarded in Attio. Watch out for: Don't nag every 24 hours. Once at 48h, once at 5 days, then escalate to a Loom or a call. The verification step matters — "did the client click the invite" beats "did I send the invite." For enterprise clients, the checklist is 3x bigger. Have a separate enterprise template. Skill file: client-onboarding

### Tip 1.3 — First-week cadence: agent-driven momentum

**What it does:** Week one isn't a sequence — it's a rhythm. The agent runs a daily morning briefing on every active onboarding client: where they are in setup, what's blocked, what their next milestone is, and the one specific thing you should do today to move them forward. Why it wins: CS people get into reactive mode. The agent gives you the proactive view. By 9am you know "Client A is blocked on the API key, Client B needs the kickoff rescheduled, Client C is ready for their first win — record their welcome Loom today." Tools: Your CRM, your project tracker (Notion DB or Linear), your calendar, your messaging channels with the client. How to wire it: 1. The agent reads every client in status = onboarding daily. 2. For each one, it pulls: setup checklist state, last message in/out, calendar status, days since deal close. 3. It scores each client on a 1-5 momentum scale (5 = humming, 1 = stalled). 4. Output is a single morning message: 3 lines per client, sorted worst-to-best, with one explicit action you should take today. Example prompt to your agent: Every morning at 8am, run my onboarding-cadence sweep. For each client with status onboarding in Attio: pull setup checklist state, last 7 days of email/Slack activity, calendar events, days since close. Score momentum 1-5. Output one Telegram message: clients sorted worst-to-best, 3 lines each — current state, what's blocking, one explicit action I should take today. Pin anything scoring 1-2 to the top with a red flag. Watch out for: Don't let the agent score on activity volume alone. A client can be quiet and happy. Cross-reference with the checklist and the kickoff plan. If you don't take the daily action, the agent should escalate. Three skipped days = it pings you with "we're about to lose momentum on X." Skill file: client-onboarding, pipeline-report

## 2. Health Monitoring

Once a client is past onboarding, the question stops being "are they set up" and becomes "are they getting value." That's a harder signal — but agents are uniquely good at it because they can watch everything: usage data, message cadence, sentiment, deliverable status.

### Tip 2.1 — Silence detection across channels

**What it does:** Your agent watches every channel you talk to a client on — email, Slack, WhatsApp, Telegram, the project tool — and flags any client who's gone silent for a threshold period. Threshold scales with deal size and stage. Why it wins: "How is X doing?" is a question CS people answer by gut feel. Gut feel misses the slow fade. A client who messaged you 4 times a week and is now messaging once every two weeks is on the slide — but if you're not measuring cadence, you don't see it until they ghost the renewal call. Tools: Your CRM, your messaging channel histories (Gmail API, Slack API, WhatsApp via Baileys, etc.), lead-status style fan-out pattern. How to wire it: 1. Per client, define an expected cadence based on their plan: high-touch = at least one substantive exchange per week, low-touch = at least one per month. 2. Daily, the agent fans out across channels, finds last-message-in and last-message-out per client, and flags any client where cadence has dropped below threshold. 3. Two flags: they went quiet (most concerning) vs you went quiet (your fault, easy fix). 4. For each flagged client, the agent loads the last 30 days of context and drafts a reengagement message in your voice — not "are you ok," more "saw [thing] in your industry, made me think of you." Example prompt to your agent: Every morning at 9am, run silence detection across all active clients. For each client, pull last-message-in and last-message-out across Gmail, Slack, WhatsApp, and the project tool. Compare to their expected cadence (tagged in Attio). Flag any client past 1.5x threshold. Split into "they went quiet" vs "I went quiet." For the "they went quiet" list, draft a re-engagement message in my voice that references something concrete from the last 30 days of their channel (a thread, a deliverable, an industry event). Stage drafts in client-followups/YYYY-MM-DD/ . Send me the list before drafts are ready. Watch out for: Not every silent client is at risk. Some clients only message at quarterly reviews. Tag those. Re-engagement messages that sound like "checking in!" are worse than no message. Force the agent to reference something specific. Volume matters less than substance. A 5-message thread last week beats 20 emoji reactions. Skill file: lead-status, stale-lead-blitz, email-followups

### Tip 2.2 — Usage signal monitoring (for SaaS / agent-as-product)

**What it does:** If your product produces usage data (logins, sessions, agent invocations, API calls), the agent monitors it per client and flags anomalies — sudden drops, sudden spikes, or flat usage that should be growing. Why it wins: Usage is the cleanest leading indicator of churn there is. A B2B SaaS account that drops to one login per week is going to churn — you have 60-90 days to save it. Most CS teams check usage dashboards once a month. The agent checks daily. Tools: Your product's usage analytics (Mixpanel, PostHog, internal logs, BigQuery), your CRM, your alerting channel. How to wire it: 1. Define usage signals per ICP tier. "Healthy" might mean: 5+ logins/week, 1+ feature deeply used, no support tickets unresolved. 2. The agent pulls usage daily, computes the 7-day rolling average per account, compares to baseline and to peer accounts of similar size. 3. Anomalies: drop > 40% week-over-week, flat for 3+ weeks, or sudden spike (could be good — could be the customer running a test before deciding to renew). 4. Daily flagged list. For each flag, the agent loads recent client comms and proposes a next action: "send the new dashboard tutorial," "schedule a quick win call," "offer a feature unlock." Example prompt to your agent: Every morning, pull last 7 days of usage data per account from PostHog. Compute rolling 7-day average per account. Compare to that account's 30-day baseline and to peer accounts in the same plan tier. Flag: drops > 40% WoW, accounts flat for 3+ weeks, or unusual spikes. For each flag, load the last 30 days of email/Slack with that account, and propose one next action (tutorial, call, feature unlock). Output: ranked list, worst signals first, with the action draft attached. Watch out for: Seasonality. Don't flag a B2B account that drops every Christmas as at-risk. Vanity metrics. "Logins" can be high while the actual value-driving action is dropping. Pick the right metric. Spikes can mean the buyer is auditing usage for renewal. Don't ignore them. Skill file: site-analytics (similar pattern), revenue-tracker

### Tip 2.3 — NPS / sentiment triggers from real conversations

**What it does:** Forget NPS surveys nobody answers. Your agent reads every conversation you have with a client across channels, scores sentiment per interaction, and computes a rolling sentiment trend per account. Sudden drops trigger an alert. Why it wins: Real-time sentiment from real conversations is a thousand times more accurate than a quarterly form. A client who writes "we're really happy with how the agents are going" three weeks running is healthy. One who suddenly switches to "this hasn't been working for us" — you find out in 24 hours instead of at the renewal call. Tools: Your email + chat archives, an LLM call per message (cheap on Haiku/Flash-class models), your CRM. How to wire it: 1. Per client, the agent reads every inbound and outbound message in the last 30 days. 2. Each message gets scored: tone (-2 to +2), engagement (1-5), urgency (1-5), explicit signals (mentions of "frustrated," "considering," "pricing," "renewal," "alternative"). 3. Rolling 14-day sentiment per account, plotted as a delta vs the prior 30 days. 4. Alert on: any explicit signal hit, sentiment delta > -1, three consecutive negative-toned messages. 5. Weekly digest: top 5 climbing accounts, top 5 declining, top 5 stable. The climbing list is your expansion list (Tip 3.x). Example prompt to your agent: Every Sunday night, run sentiment-trend per active client. Pull last 60 days of messages across all channels. Score each message: tone (-2..+2), engagement (1-5), urgency (1-5), explicit signals (frustrated, considering, pricing, renewal, alternative). Compute 14-day rolling sentiment vs the prior 30 days. Output 3 lists: climbing (sentiment up + no negative explicit signals), declining (sentiment down OR any explicit negative signal), stable. Send Monday morning. Flag any client crossing into declining with red. Watch out for: Sentiment models miss sarcasm and inside jokes. Trust direction, not absolute scores. A short message isn't a negative message. Normalize for length. Don't act on a single negative interaction. Trend matters more than any single signal. Skill file: lead-status, fathom-transcripts

## 3. Renewal & Expansion

Most CS teams treat renewal as a date on a calendar — 30 days out, send the renewal email, hope for the best. That's how you find out a client churned. The agent's job: surface the renewal and expansion conversation 90 days out, and make sure you walk into it with every piece of context the client could possibly throw at you.

### Tip 3.1 — Renewal radar (90/60/30/15 day cron)

**What it does:** Per client, your agent runs a renewal-readiness check at T-90, T-60, T-30, and T-15. Each check pulls usage, sentiment, deliverable status, and outstanding issues, and produces a one-page "where this account stands" brief. T-90 starts the save plan if anything's yellow. T-30 drafts the renewal call invite. Why it wins: Renewals don't get saved 7 days out. They get saved at T-90 when you still have time to deliver a quick win, fix the broken integration, or reset expectations. Most CS teams find out at T-30. The agent gives you the 60 extra days. Tools: Your CRM (contract end date field), your usage and sentiment trackers from Section 2, your calendar for the renewal call. How to wire it: 1. Cron daily: agent reads every active contract, computes days-to-renewal. 2. Triggers at 90, 60, 30, 15 days. Each trigger fires a full account brief: usage trend, sentiment trend, deliverable status, recent escalations, expansion signals. 3. T-90 brief includes a save-plan recommendation if any signal is yellow. 4. T-30 brief includes the draft renewal-call invite in your voice, with a custom recap of what's been delivered. 5. T-15 brief is the "shit, we still haven't booked it" alarm. Example prompt to your agent: Daily, compute days-to-renewal per active contract. At T-90, generate the full account brief: usage trend (Tip 2.2 output), sentiment trend (Tip 2.3 output), deliverable status, escalations, expansion candidates. If any signal is yellow, propose a save plan: 1-2 concrete moves over the next 60 days. At T-60 update the brief. At T-30, draft the renewal-call invite in my voice with a 5-bullet recap of what's been delivered, and stage it. At T-15, if the renewal call isn't booked, alarm me on Telegram and draft a more direct nudge. Watch out for: Auto-renewal contracts are different. Run the same checks but the action is "preempt the cancel email," not "book the call." T-90 yellow signals require a plan, not a panic call. Don't let the agent draft a "we need to talk" email.

Expansion lives next to renewal — see 3.2. Skill file: pipeline-report, stale-lead-blitz, post-call-autopilot

### Tip 3.2 — Expansion signal hunter

**What it does:** Your agent reads every active client interaction looking for explicit expansion signals — "do you also do X," "we're hiring more people," "the team across the hall is looking at this too," "can this also handle Y." It flags them, drafts a soft introduction of the relevant offering, and queues it for your next conversation. Why it wins: Clients ask the expansion question. Most CS people miss it because they're focused on solving the current ticket. The agent catches every "by the way" comment and puts it in front of you with a draft response. Tools: Your conversation archives, your offering catalog (one paragraph per product/service tagged with trigger keywords), your CRM. How to wire it: 1. Build a expansion-triggers.json config: per offering, the language clients use when they're hinting they want it. 2. Daily, the agent scans new messages across all channels for trigger matches. 3. For each hit, it loads the surrounding context, the relevant offering one-pager, and drafts a soft response: "you mentioned X — we actually do this for a few clients already. Want me to send a quick overview, or are you just thinking out loud?" 4. Drafts queue with your next reply to that client. Never auto-sent. Example prompt to your agent: Build an expansion-radar skill. Read expansion-triggers.json (offering → trigger keywords). Every day at 4pm, scan the last 24h of client messages across all channels. For each trigger hit, load: the surrounding 5-message context, the matching offering one-pager from offerings/ , and the client's deal history. Draft a soft response in my voice. Queue it as a suggested reply on my next message to that client. Send me a daily summary of triggers caught. Watch out for: Don't pitch on the same thread where the client is asking for help. Wait one beat. "We're growing" is a softer signal than "we're hiring 10 SDRs next quarter." Score the heat. The offering one-pagers need to be tight. If they're 3 pages long, the agent will quote too much in the draft.

### Skill file: outreach-drafter, deal-accelerator

### Tip 3.3 — Save-call invite drafter

**What it does:** When the renewal radar (Tip 3.1) flags a client as trending toward churn, your agent doesn't just notify you — it drafts the actual save-call invite. The invite references specific recent friction, proposes a concrete agenda, and offers two time slots from your real calendar. Why it wins: Half-drafted save calls die in your drafts folder. A ready-to-send invite with real context lowers the friction enough that you actually send it the same day. Tools: Your email voice skill, your calendar via Google Calendar API or gws , the account brief from Tip 3.1. How to wire it: 1. When 3.1 flags an account as yellow or red, the agent drafts the invite. 2. Invite includes: a one-line warm opener, a specific reference to the friction ("noticed we haven't shipped X yet"), a proposed 30-minute agenda (3 bullets), and two real calendar slots in the next 5 business days. 3. Stage in your drafts. You tweak and send. Example prompt to your agent: When the renewal radar flags any account yellow or red, draft a save-call invite using my email-voice skill. Include: 1-line warm opener, specific reference to the friction (pulled from the brief), 3-bullet 30-min agenda, two calendar slots in the next 5 business days (pulled live from my Google Calendar — avoid blocks). Stage in my Gmail drafts and ping me on Telegram with a link. Watch out for: Don't over-acknowledge the friction. One sentence, then forward-looking. The calendar slots have to be real and current. Stale slots make you look sloppy. For larger accounts, the invite should propose a video call with a specific exec — not a generic "let's chat." Skill file: email-followups, meeting-prep, calendar-event

## 4. Done-For-You Async Check-Ins

The highest-leverage move in modern CS: replace the recurring "status call" with an async report the client can watch on their schedule. Done well, it makes the client feel more cared

for than the meeting did — because they can rewatch it, share it with their team, and you didn't take 30 minutes from each side.

### Tip 4.1 — Weekly Loom-style report (auto-generated)

**What it does:** Every Friday, your agent generates a per-client weekly report: a 3-5 minute video that walks through what got done this week, the metric movement, anything that's blocked, and what's coming next week. The agent assembles the script from the project log and your delivery channels. You record or auto-narrate. Why it wins: Clients who get a tangible weekly artifact stay 2-3x longer than clients who don't. It anchors the value you're delivering. It also reduces the inbound "what's the status" questions to zero. Tools: Your project tracker (Notion DB / Linear / Asana), your usage and metric dashboards, a screen recording tool (Loom or Tella) with auto-script, optionally a voice clone if you've trained one. How to wire it: 1. Friday morning, the agent assembles the per-client report data: what shipped this week, metric movements, open blockers, next week's plan. 2. It writes a script in your voice with timestamps and visual cues ("now show the dashboard with the conversion graph"). 3. Option A: you record the Loom in 5 minutes per client following the script. 4. Option B: it auto-narrates via a voice clone + screen capture stitched together with stills and screenshots. 5. Report goes out Friday afternoon with a short email summary and the video link. Example prompt to your agent: Every Friday at 10am, generate per-client weekly reports for all active accounts. For each client, pull: shipped items from Linear, metric movements from analytics, open blockers, next week's plan. Write a 3-5 min Loom script in my voice: 30s warm opener with their name, 90s what shipped, 60s metric movement with screenshots referenced, 30s what's blocked + what we need, 60s next week. Stage scripts at clientreports/YYYY-MM-DD/<slug>.md with all the screenshots linked. I'll record. Then you email each one out Friday 3pm with the Loom link and a 3-bullet summary. Watch out for: Don't bury the lead. Open with the biggest win of the week, not "this week we worked on..." Always include a forward-looking element. Reports that only look backward feel like billing.

Voice clones still sound a little off to people who know you. If the client is high-value, record it yourself. Skill file: consultation-recap, post-call-autopilot, email-followups

### Tip 4.2 — Async quarterly business review (QBR)

**What it does:** Same pattern as 4.1 but quarterly and deeper. The agent assembles a full QBR doc: results vs goals, the story of the quarter, what worked, what didn't, expansion ideas, and the proposed plan for next quarter. You record a 15-20 minute walkthrough video. Client gets the doc + the video. Why it wins: Live QBRs eat half a day on both sides and get scheduled three weeks late because nobody can find the time. The async version lands on day one of the quarter, gets watched by the actual decision-makers (not just your champion), and forms the basis of every renewal conversation 90 days later. Tools: Same as 4.1, plus your slide tool (or a Google Doc template), and your data warehouse if you have one. How to wire it: 1. Define a QBR template: results vs goals (with numbers), the narrative, lessons, expansion proposals, next quarter plan. 2. End of quarter, the agent fills the template per client using all available data. 3. It drafts a 15-20 min walkthrough script. 4. You record the walkthrough — or, for accounts under a certain size, auto-narrate. 5. Send doc + video on day 1 of the new quarter, with a 30-min "questions about the QBR" booking link. Example prompt to your agent: Build a qbr-generator skill. Inputs: client slug, quarter end date. Pull: shipped deliverables, metric movements vs the goals tagged in Attio, all major comms threads from the quarter, support tickets, sentiment trend. Fill the QBR template at templates/qbr.md . Write a 15-20min walkthrough script. Run for every active client in the last 3 business days of the quarter. Output: per-client doc and script. I record, you email it day 1 of the new quarter with a 30-min booking link. Watch out for: A QBR with bad numbers is still a QBR. Don't let the agent gloss over misses — name them and propose the fix. Include the client's name and role explicitly in the opener. Generic QBRs feel generic.

If the data isn't there, the QBR isn't there. Set up the tracking before the quarter, not at the end. Skill file: consultation-recap, pdf-proposal (same artifact pattern)

### Tip 4.3 — "Surprise win" pings

**What it does:** When a client hits a milestone (first 10k revenue from your agent, 100th lead booked, 6-month anniversary, etc.), your agent catches it and either drafts a quick congrats note or, for bigger milestones, kicks off a small gesture (handwritten card via API, a custom Loom, a shoutout post if they're OK with it). Why it wins: The unexpected ping at the unexpected moment is the highest-ROI thing in client retention. Clients don't remember the scheduled QBR. They remember the day you noticed they hit 100 deals. Tools: Your milestone tracker (numeric fields in the CRM or product DB), email/Slack channels, optionally a handwritten-card-as-a-service API (Postable, Felt, etc.). How to wire it: 1. Define a milestones.json : per client or per offering, what counts as a milestone and what the response should be (note / Loom / card). 2. Daily, the agent checks every client's metrics against milestone thresholds. 3. On hit, the agent drafts the appropriate response and queues it. 4. You approve / personalize / send. Example prompt to your agent: Daily at 11am, check every active client's product metrics against milestones.json thresholds (e.g. 100th lead, 6-month anniversary, $50k attributable revenue). On hit, draft the milestone response per the config: short note, Loom script, or handwritten card. For the card type, queue with the Felt API but don't send until I approve. For Loom scripts, stage them at milestones/YYYY-MM-DD/ . Ping me with the daily list. Watch out for: Don't fake intimacy. A milestone ping that sounds templated is worse than no ping. Roll the celebration into something useful — "hit 100 leads, here's a Loom on the 3 patterns I see in the ones that converted." Don't ping every 100. Pick the milestones that actually matter to the client. Skill file: client-onboarding, email-followups

### How it all stacks

CS works because every layer reinforces the next. Onboarding lands the client. Health
monitoring catches drift. Renewal radar gives you 90 days to save what drifts. Async checkins keep the value visible. Surprise wins drive the loyalty that makes renewal a formality.
Install order:
1. Onboarding sequence (Tip 1.1) and the setup checklist (Tip 1.2). Without these, you're starting every relationship in a hole.
2. Silence detection (Tip 2.1). The single highest-leverage tip in the section. You'll catch your first save inside 30 days of installing.
3. Renewal radar (Tip 3.1). Once you have a few clients past the 90-day mark, this is what protects MRR.
4. Weekly async reports (Tip 4.1). This is what turns clients into 24-month accounts instead of 6-month accounts.
5. Sentiment trend (Tip 2.3) and expansion radar (Tip 3.2). These are the expansion engine — install once the basics are solid.
6. QBR auto-gen (Tip 4.2) and milestone pings (Tip 4.3). Polish layer. High impact per touch, lower frequency. Don't try to ship all 10 tips in a month. Start with onboarding and silence detection. Get those bulletproof. Everything else compounds from there.

### Product & Engineering