--- name: good_morning_cfo description: > Daily executive briefing for the office of the CFO. Covers revenue performance vs forecast, top account movers, expense anomalies vs budget, consumption trends, and external market signals — delivered as a single consolidated brief under 600 words. Triggers: good morning, daily briefing, morning briefing, CFO update, start my day, what should I know today, daily revenue update, morning revenue update. dehydrated: true data_domain: "finance daily executive briefing, revenue forecast vs actuals, account-level revenue trends, expense vs budget anomalies, consumption usage" mcp_optional: [web_search] created_date: 2026-05-19 version: 1.1.0 license: Snowflake Skills License. See LICENSE.txt for complete terms type: snowflake language: en status: Published --- # Good Morning CFO (Dehydrated) ## Intent A 5-minute daily briefing for the CFO and FP&A leadership. Surfaces revenue performance against forecast, top accounts driving variance, expense anomalies vs budget, consumption trends, and a brief market scan — in a single consolidated brief. Designed to replace the morning routine of pulling three dashboards and scanning Slack channels. Not for: deep root-cause analysis (use `revenue_analysis` or `budget_vs_actuals`), dashboard creation, transactional data lookups, or scenario modeling (use `scenario_planning`). ## Behavior ### Trigger Detection Activate on: "good morning", "daily briefing", "morning briefing", "CFO update", "start my day", "what should I know today", "morning revenue update". If the prompt is a generic "good morning" with no finance context, confirm via `ask_user_question` before pulling data — this skill hits multiple tables per run. Do NOT activate for: - Revenue deep dives (route to `revenue_analysis`) - Expense variance analysis (route to `budget_vs_actuals`) - Forecast comparison across snapshots (route to `forecast_vs_forecast`) ### Workflow **Run all four streams in parallel — do not stream partial results. Wait for all to complete, then deliver the full brief in one response.** | Stream | Data Source | What It Computes | |---|---|---| | A — Revenue pulse | `finance.revenue.forecast_daily` | Yesterday actual vs forecast; MTD and QTD pacing | | B — Top movers | `finance.revenue.account_forecast` | Top 5 accounts driving positive variance; top 5 driving negative | | C — Expense anomalies | `finance.budget.expense` | Lines where \|actual vs budget\| > threshold; flag favorable and unfavorable | | D — External signals | web_search | Top 3 macro or industry signals from the last 24 hours | **Stream E — Consumption pulse (optional):** If `finance.consumption.daily_usage` resolves, include a fifth section. If LOW confidence, skip silently with no mention. **Time window defaults:** - Revenue pulse: yesterday vs prior business day; MTD vs MTD prior year - Use the latest day that has actuals. If yesterday has not landed, say "No data yet for — last available: " and run the pulse on the last available day. - Compare actuals to a forecast version created **before** the day being measured. In a table that holds many forecast versions, the latest version often equals actuals for past days, which makes every day look on forecast. Say which version is the baseline. - Show MTD and QTD against the same pre-dated forecast baseline and against prior year. - Top movers: most recent week ending date vs prior week. If a mover's change is more than ~90% down (or the account goes to near zero), describe it as a possible contract end, migration or data gap to verify, not as an ordinary trend, and say what was checked. - Expense anomalies: most recent closed month vs budget. Before ranking, remove subtotal and rollup rows and keep revenue lines separate from expense lines (see the cleanup step in `budget_vs_actuals`), otherwise totals and parent rows crowd out the real line items. - External signals: last 24 hours If the user specifies a different scope or time window, apply it to all streams. **After delivery:** Ask "Anything here you want to drill into?" — do not proactively run follow-up queries without user confirmation. Offer these follow-ups: a segment or account drill-down, a flagged expense line, or a short email draft of the brief for the user to review before anything is sent. ### Output Format Output rules: - Begin the response with the greeting line. Do not narrate the data pulls or print any text before it. - Label fiscal periods as `FY-Q` using the customer's configured fiscal calendar. - For revenue, show actual against forecast **and** against the prior-year period when a prior-year comparison exists, so a beat against forecast is not read in isolation. - Expense anomalies use the same tiered materiality as `budget_vs_actuals`. If the expense data includes a policy-exception flag (for example out-of-policy travel and expense lines), list those separately beneath the anomalies table. ``` # Good Morning, ** — ** ## Revenue Performance - Yesterday: $XXM actual vs $XXM forecast (+/-$XXM, +/-X.X%) - MTD: $XXM actual, pacing - QTD: $XXM actual vs $XXM target (%) ## Top Movers (WoW) **Driving positive variance:** 1. — +$XXM () 2. ... **Driving negative variance:** 1. — -$XXM 2. ... ## Expense Anomalies | Line Item | Budget | Actual | Variance | Flag | | | $XXM | $XXM | +/-$XXM (+/-X.X%) | FAV / UNFAV | <2-sentence commentary on the largest flagged line> ## Consumption Pulse [include only if data available] - WoW usage: +X.X% - Top growing accounts: - Watch list (declining): ## External Signals - [] - [] --- Hydration: , , ``` ### Confidence Handling | Confidence | Behavior | |---|---| | HIGH (Streams A, B, C all resolved) | Run all streams; full brief | | MEDIUM (any stream has column ambiguity) | Run all streams; prefix brief with: "Some bindings are MEDIUM confidence — verify hydration_hints.yaml" | | LOW on non-critical (Stream C or E missing) | Skip that section; note "data unavailable"; produce brief from remaining streams | | LOW on critical (Stream A — forecast_daily — missing) | Block. Output: "Daily briefing requires a revenue forecast binding. Add an entry under finance.revenue.forecast_daily in hydration_hints.yaml and re-run." | ### Guardrails - Never expose internal system identifiers or raw IDs in output — use account names. - Never assume a fiscal calendar. Read from plugin config (`feb_start`, `jan_start`, or `custom`). - If no data has landed for the requested window, surface this explicitly: "No data yet for — last available: ." - Currency: respect the `currency` column. Display in dominant currency; group by currency if multiple present. - Web search: only include sources from the last 24 hours. If no relevant results, skip the section silently. - Expense anomalies: apply a materiality threshold (suggest >5% OR >$50K absolute). Do not flag immaterial lines. Confirm the thresholds with the user rather than hardcoding them. - Keep the full brief under 600 words. ### Error Handling | Scenario | Action | |---|---| | Critical binding (Stream A) fails | Block briefing; output setup instructions | | Query timeout | Retry once with a 30-day window; if still fails, omit that section | | Web search returns nothing | Skip External Signals silently | | User asks for a different scope mid-briefing | Re-run with new scope; do not merge scopes | | Expense binding missing | Skip anomalies section; note "expense data not available" | ## Context Queries - A table or semantic view at the (date, segment) grain with actual revenue, forecast revenue, and a currency code — used for the Revenue Pulse section (headline beat/miss, MTD pacing, QTD attainment). If a semantic view is available, complex follow-up questions route through Cortex Analyst. *(critical)* - A table at the (week_ending_date, account_id) grain with the current week's forecast amount per account. The skill computes WoW change via LAG() — no pre-computed delta column required. Used for the Top Movers section. *(critical)* - A table at the (period, line_item_id) grain with budget amount and actual amount per line item. Used for the Expense Anomalies section. Materiality threshold applied at runtime. *(non-critical — section skipped if absent)* - (Optional) A table at the (usage_date, account_id) grain with a usage or units column for the Consumption Pulse section. Skill computes WoW change in SQL. Skipped if absent.