Prepare Customer Interview

Slack MCP (user-slack)

Mixpanel MCP (user-mixpanel)

Web

Gong MCP (user-gong)

Monday API MCP (user-monday-api-mcp)

Workflow

Step 1 -- Read the data guide

Read your data guide for current table schemas, join keys, and known gotchas.

Step 2 -- Resolve the account

Based on the identifier type:

Account ID: Use directly as PULSE_ACCOUNT_ID.

Company name: Query Snowflake to find matching accounts:

SELECT PULSE_ACCOUNT_ID, PRODUCT, FIRST_PRODUCT_INSTALL_DATE,
       IS_PAYING_PRODUCT, ACCOUNT_INFRA_REGION
FROM BIGBRAIN.L4.DIM_ACCOUNTS_PRODUCTS_METRICS
WHERE PRODUCT = 'marketing'
  AND LOWER(PULSE_ACCOUNT_ID::VARCHAR) LIKE '%{search_term}%'

If no results, use data-expert-agent to search across account tables for the company name.

Email/domain: Extract the domain and search. Also use the domain to construct the website URL (https://{domain}).

If multiple accounts match, present them to the user and ask which one to research.

Step 3 -- Pull product usage data

Query campaigns activity for the resolved account:

SELECT c.ID, c.NAME, c.TYPE, c.STATUS, c.SENTON, c.CREATEDAT
FROM BIGBRAIN.STG.STG_MARKETING_CHANNELS_CAMPAIGNS c
WHERE c.ACCOUNTID = {account_id}
  AND c.SOURCE != 'IL'
ORDER BY c.CREATEDAT DESC

Then pull campaign performance metrics:

SELECT s.CAMPAIGNID, s.TOTALSENT, s.TOTALDELIVERED,
       s.TOTALOPENS, s.UNIQUEOPENS, s.TOTALCLICKS, s.UNIQUECLICKS
FROM BIGBRAIN.L1.MARKETING_CHANNELS_EMAILCAMPAIGNANALYTICSSUMMARY_US s
WHERE s.ACCOUNTID = {account_id}

Repeat for EU and AU regions if the account's region warrants it.

Summarise:

• Total campaigns created vs sent
• Campaign types used
• Send volumes and frequency (last 30d, last 90d, all time)
• Performance: open rate, click rate, bounce rate
• Most recent campaign date

Step 4 -- Pull financial data

SELECT DAY, PLAN_TIER, PLAN_PERIOD, SEATS, BILLING_ARR, PRODUCT_ARR,
       PRODUCT_ARR_DAILY_CHANGE, PRODUCT_ARR_DAILY_CHANGE_TYPE,
       DAU, MAPP, IS_PAYING_PRODUCT, IS_IN_TRIAL
FROM BIGBRAIN.L4.FACT_ACCOUNTS_PRODUCTS_METRICS_DAILY
WHERE PULSE_ACCOUNT_ID = {account_id}
  AND PRODUCT = 'marketing'
ORDER BY DAY DESC
LIMIT 90

Also pull subscription lifecycle:

SELECT EVENT_TYPE, EVENT_TIMESTAMP, CONTRACT_STATUS,
       CANCELLATION_REASON, CANCELLATION_REASON_DESC,
       ARR, NEXT_ARR, PLAN_TIER, PERIOD,
       RENEWAL_STATUS, CANCEL_ON_RENEWAL_IND
FROM BIGBRAIN.L4.DIM_ACCOUNT_SUBSCRIPTION_CONTRACTS
WHERE PULSE_ACCOUNT_ID = {account_id}
ORDER BY EVENT_TIMESTAMP DESC

Summarise:

• Current plan tier, period, seats, ARR
• Billing history (upgrades, downgrades, churn events)
• Trial status and dates
• Cancellation signals (cancel-on-renewal, churn reason if any)
• MAPP and DAU trend (last 30 days)

Step 5 -- Pull other monday products

SELECT PRODUCT, FIRST_PRODUCT_INSTALL_DATE, IS_PAYING_PRODUCT,
       FIRST_BILLING_ARR_DATE, FIRST_WEEK_DAU_7,
       IS_SUCCESSFUL_SETUP, CHANNEL, SIGNUP_FLOW
FROM BIGBRAIN.L4.DIM_ACCOUNTS_PRODUCTS_METRICS
WHERE PULSE_ACCOUNT_ID = {account_id}
ORDER BY IS_PAYING_PRODUCT DESC, FIRST_PRODUCT_INSTALL_DATE

Summarise:

• Which monday products are installed and active
• Which are paid vs free
• Tenure per product
• Engagement level (first week DAU as proxy)

Step 6 -- Search for customer feedback (Slack)

Search campaigns-stats and other relevant channels for mentions of the customer:

1. Use slack_search_public with the company name as the query
2. Also search with the account ID if available
3. Check campaigns-stats, ask-campaigns, and campaigns-churns channels
4. For relevant results, read the full thread with slack_read_thread

Summarise:

• Feedback themes (positive, negative, feature requests)
• Support issues raised
• Direct quotes worth noting
• Links to relevant Slack threads

Step 7 -- Pull Gong call history

Search for previous calls with this customer using the company name or known contacts.

Workflow:

1. list_calls with a broad date range (last 6 months) to find relevant calls
2. If available, filter by searching call titles or metadata containing the company name
3. get_call_summary for each matching call (prioritise the 3-5 most recent)
4. For calls with interesting summaries, use get_call to retrieve the call URL
5. Pull the transcript with get_call_transcript to find exact quotes worth referencing in the interview
6. Construct a deep link for each notable quote: {call_url}#t={timestamp_in_seconds} (timestamp comes from the transcript's timing data)

Summarise:

• Number of calls on record and date range
• Key topics discussed across calls (recurring themes)
• Action items or commitments made in previous conversations
• Pain points or feature requests mentioned on calls
• Sentiment trajectory (are calls getting more positive or negative over time?)
• Competitor mentions in call discussions
• Notable verbatim quotes with Gong deep links (so the user can listen before the interview)

Note: If Gong is not configured or no calls are found for this customer, note it in the output and move on.

Step 8 -- Search for support tickets (best effort)

Use user-monday-api-mcp search tool to look for the company name across monday boards. If a support or tickets board exists, pull relevant items.

If nothing is found, note: "No dedicated support ticket system accessible. Slack search was used as the primary source for support-related signals."

Step 9 -- Pull company website

Use WebFetch to retrieve the customer's website. Extract:

• What the company does (industry, product/service)
• Company size indicators (team page, about page)
• Target market
• Any marketing-related context (do they have a blog, newsletter, social presence?)

If the website URL is unknown, ask the user to provide it.

Step 10 -- Compile and generate output

Short summary (in conversation)

Present 5-7 bullet points covering:

• Who they are (company, industry, size)
• How they use campaigns (volume, types, performance)
• Financial status (plan, ARR, tenure, risk signals)
• Other monday products in use
• Notable feedback or support signals
• Previous call history and key themes from Gong (if available)
• One or two suggested areas to probe in the interview

HTML dossier

Generate an HTML file using the routine template at skills/system/routine-html-template.md. Read that file for the full CSS and component patterns.

Save to: your output folder/customer-interview-prep-{company-slug}-YYYY-MM-DD.html Open with: open command

HTML Structure

<div class="header">
  <h1>Customer Interview Prep</h1>
  <div class="subtitle">{Company Name} · {date}</div>
</div>

<!-- Company Overview -->
<div class="section">
  <div class="section-title">Company Overview</div>
  <div class="card">
    <div class="card-body">[What the company does, industry, size, target market]</div>
  </div>
</div>

<!-- Account Snapshot -->
<div class="section">
  <div class="section-title">Account Snapshot</div>
  <div class="metric-grid">
    <div class="metric-card">
      <div class="metric-value">[tier]</div>
      <div class="metric-label">Plan</div>
    </div>
    <div class="metric-card">
      <div class="metric-value">$[N]</div>
      <div class="metric-label">ARR</div>
    </div>
    <div class="metric-card">
      <div class="metric-value">[N]</div>
      <div class="metric-label">Seats</div>
    </div>
    <div class="metric-card">
      <div class="metric-value">[N] months</div>
      <div class="metric-label">Tenure</div>
    </div>
    <div class="metric-card">
      <div class="metric-value">[region]</div>
      <div class="metric-label">Region</div>
    </div>
  </div>
</div>

<!-- Product Usage -->
<div class="section">
  <div class="section-title">Campaigns Usage</div>
  <div class="metric-grid">
    <div class="metric-card">
      <div class="metric-value">[N]</div>
      <div class="metric-label">Campaigns created</div>
    </div>
    <div class="metric-card">
      <div class="metric-value">[N]</div>
      <div class="metric-label">Campaigns sent</div>
    </div>
    <div class="metric-card">
      <div class="metric-value">[N]</div>
      <div class="metric-label">Total emails sent</div>
    </div>
    <div class="metric-card">
      <div class="metric-value">[%]</div>
      <div class="metric-label">Avg open rate</div>
    </div>
    <div class="metric-card">
      <div class="metric-value">[%]</div>
      <div class="metric-label">Avg click rate</div>
    </div>
  </div>
  <table>
    <thead><tr><th>Period</th><th>Campaigns sent</th><th>Emails sent</th><th>Open rate</th><th>Click rate</th></tr></thead>
    <tbody>
      <tr><td>Last 30 days</td><td>[N]</td><td>[N]</td><td>[%]</td><td>[%]</td></tr>
      <tr><td>Last 90 days</td><td>[N]</td><td>[N]</td><td>[%]</td><td>[%]</td></tr>
      <tr><td>All time</td><td>[N]</td><td>[N]</td><td>[%]</td><td>[%]</td></tr>
    </tbody>
  </table>
  <div class="card">
    <div class="card-title">Recent campaigns</div>
    <div class="card-body">[List of last 5 campaigns with name, date, type, status]</div>
  </div>
</div>

<!-- Other Monday Products -->
<div class="section">
  <div class="section-title">Other Monday Products</div>
  <table>
    <thead><tr><th>Product</th><th>Installed</th><th>Paying</th><th>First week DAU</th></tr></thead>
    <tbody>
      <tr><td>[product]</td><td>[date]</td><td>[yes/no]</td><td>[N]</td></tr>
    </tbody>
  </table>
</div>

<!-- Financial History -->
<div class="section">
  <div class="section-title">Financial History</div>
  <div class="card [success/warning/urgent]">
    <div class="card-title">[Current status summary]</div>
    <div class="card-body">[Billing events timeline: upgrades, downgrades, churn signals]</div>
  </div>
</div>

<!-- Customer Voice -->
<div class="section">
  <div class="section-title">Customer Voice</div>
  <!-- One card per feedback source or theme -->
  <div class="card highlight">
    <div class="card-label highlight">campaigns-stats</div>
    <div class="card-body">[Feedback summary with quotes]</div>
    <div class="card-meta">[Link to Slack thread]</div>
  </div>
  <div class="card">
    <div class="card-label">ask-campaigns</div>
    <div class="card-body">[Support questions or issues raised]</div>
    <div class="card-meta">[Link to Slack thread]</div>
  </div>
</div>

<!-- Previous Calls (Gong) -->
<div class="section">
  <div class="section-title">Previous Calls · Gong</div>
  <div class="metric-grid">
    <div class="metric-card">
      <div class="metric-value">[N]</div>
      <div class="metric-label">Calls on record</div>
    </div>
    <div class="metric-card">
      <div class="metric-value">[date]</div>
      <div class="metric-label">Most recent call</div>
    </div>
  </div>
  <!-- One card per notable call -->
  <div class="card highlight">
    <div class="card-label highlight">[Call date] · [Call title]</div>
    <div class="card-body">[Key points from call summary: topics, pain points, action items, competitor mentions]</div>
  </div>
  <!-- Recurring themes across calls -->
  <div class="card warning">
    <div class="card-label warning">Recurring themes</div>
    <div class="card-body">[Themes that appear across multiple calls: feature requests, objections, frustrations]</div>
  </div>
  <!-- Notable quotes with deep links -->
  <div class="card">
    <div class="card-label">Notable quotes</div>
    <div class="card-body">
      <blockquote>"[Verbatim quote from customer]"<br>
        <span class="quiet">— [Speaker], <a href="[call_url]#t=[timestamp_seconds]">[Call title, date]</a></span>
      </blockquote>
      <blockquote>"[Another quote]"<br>
        <span class="quiet">— [Speaker], <a href="[call_url]#t=[timestamp_seconds]">[Call title, date]</a></span>
      </blockquote>
    </div>
  </div>
  <p class="quiet">If no Gong calls found: "No previous calls found in Gong for this customer."</p>
</div>

<!-- Suggested Interview Questions -->
<div class="section">
  <div class="section-title">Suggested Interview Questions</div>
  <ol class="priority-list">
    <li>
      <div class="card-title">[Question based on data gap or interesting signal]</div>
      <div class="card-body">[Why this question matters, what data prompted it]</div>
    </li>
  </ol>
</div>

Guidelines

• Lead with the most important signals, not exhaustive data dumps
• Be specific with numbers: "sent 47 campaigns in the last 90 days" not "active user"
• If a data source returns nothing, include an empty-state note in the HTML rather than omitting the section
• Flag data quality issues honestly (region gaps, missing tables, small samples)
• Suggested interview questions should be grounded in the data, not generic
• Keep the summary scannable. the user should understand the picture in 30 seconds
• Preserve Slack thread links so the user can jump to the original conversation

Change Log