This content is copy pasted from a proprietary Finance Agent (Max bought the license): https://www.notion.so/Finance-AI-Agent-Infrastructure-2d69e6cd0866808bba72d038b20ff791
Core Architecture & Capabilities The Finance AI Agent operates as a multi-stage intelligence pipeline that transforms raw financial data into board-ready presentations.
System Overview [Financial Data Sources] ↓ [Business Context Generator] (Claude Sonnet 4) ↓ [Data Extraction & Normalization] ↓ [CFO-Level Variance Analysis] (Claude Sonnet 4) ↓ [Chart Generation] (Nano Banana Pro) ↓ [Presentation Assembly] (Gamma) ↓ [Board-Ready Monthly Report] (8 minutes)
Stage 1: Data Collection Form-Based Data Input The system collects five critical inputs: Company Name Organizational identifier for presentation branding Business Context Market positioning and competitive landscape Product portfolio details Seasonality patterns Strategic initiatives Current Month Financial Data (File Upload) Revenue by product line Operating expenses by category Customer acquisition metrics Churn and retention data Cash flow statement Forecasted Financial Data (File Upload) Budget vs actual comparison targets Forward-looking projections Scenario modeling data New Initiatives (Optional) Recent product launches Market expansion activities Organizational changes Competitive responses
Stage 2: Business Context Profile Generation Structured Context Engineering Claude Sonnet 4 transforms raw business description into structured JSON profile containing: Company Profile Industry classification Company stage and employee count Target market segment Market growth rate Competitive Landscape Market position Top 3 competitors Primary competitive advantage Win rate vs main competitor Product Portfolio Core product information Pricing model (starter/mid/enterprise tiers) Professional services revenue mix Value Proposition Primary differentiator Target buyer persona Main customer pain point solved Business Patterns Average sales cycle length Contract duration norms Seasonal revenue patterns (Q1-Q4 indices) Peak and trough months Strategic Initiatives Current initiative tracking (2x parallel) Monthly incremental spend Expected revenue impact Implementation status Recent product launches Market Conditions Macro sentiment analysis Customer budget environment Industry trends (positive/negative) Recent competitor actions and impact Why This Matters: This structured context enables the AI to generate intelligent variance commentary that explains "why" numbers changed—not just "what" changed. Example: Without Context: "Revenue decreased 8%" With Context: "Revenue decreased 8% primarily due to expected Q1 seasonality (historical Q1 index: 85). This aligns with our B2B enterprise sales cycle where contracts typically renew in Q4, creating predictable Q1 softness. The decline is within expected parameters and positions us for Q2 recovery as pipeline converts."
Stage 3: Financial Data Extraction Automated File Processing The system extracts structured data from uploaded files (CSV, Excel, PDF): Current Month Extraction: Revenue metrics (by product, segment, geography) Expense breakdown (COGS, SG&A, R&D, Sales & Marketing) Customer metrics (CAC, LTV, Churn Rate, NRR) Cash flow components Forecasted Data Extraction: Budget targets for comparison Forward-looking projections Scenario modeling inputs Data Normalization: Standardizes formats across different accounting systems Aggregates data into analysis-ready structure Validates data completeness and flags anomalies
Stage 4: CFO-Level Variance Analysis Multi-Dimensional Intelligence Engine Claude Sonnet 4 executes six parallel analysis streams:
Stage 5: Visual Intelligence Generation Nano Banana Pro Integration The system generates 8-12 professional charts automatically: Revenue Visualizations Revenue Trend Analysis (Multi-line chart) Current vs Prior Month Budget vs Actual Year-over-Year comparison Product Mix Evolution (Stacked bar chart) Revenue by product line Segment contribution changes Customer Segment Performance (Grouped bar chart) New vs Expansion vs Renewal Segment-level growth rates Expense Visualizations Operating Expense Breakdown (Pie chart with trend) Category distribution Month-over-month changes Efficiency Metrics Dashboard (Combo chart) Gross margin % Operating margin % EBITDA margin % Customer Metrics Visualizations CAC & LTV Trends (Dual-axis line chart) CAC evolution LTV tracking LTV:CAC ratio Retention Analysis (Cohort retention curve) Monthly cohort tracking NRR trend Churn Analysis (Waterfall chart) Customer additions vs losses Churn drivers breakdown Forward-Looking Visualizations Quarterly Forecast (Projection with confidence intervals) Revenue forecast Expense trajectory Profitability path Cash Runway Analysis (Burn rate projection) Current cash position Monthly burn rate Runway months remaining Chart Generation Process: AI analyzes financial data patterns Determines optimal visualization types Generates chart specifications (JSON) Nano Banana Pro renders professional charts Charts uploaded and linked to presentation
Stage 6: Presentation Assembly Gamma Automation The system creates a complete board-ready presentation: Slide Structure: Slide 1: Executive Summary Company name and reporting period 3-paragraph executive overview Key highlights (3-4 bullets) Strategic implications Slide 2: Revenue Performance Revenue trend chart Product mix visualization Key variance commentary Action items Slide 3: Profitability Analysis Margin trends (Gross, Operating, EBITDA) Efficiency metrics dashboard Cost structure breakdown Optimization opportunities Slide 4: Customer Metrics CAC & LTV evolution Retention analysis Churn breakdown Cohort performance Slide 5: Expense Deep Dive Operating expense breakdown Category-level variance Investment justification Cost containment wins Slide 6: Forward-Looking Outlook Quarterly forecast chart Risk factors and mitigation Strategic recommendations Next quarter priorities Slide 7: Appendix Detailed financial tables Supporting data and calculations Methodology notes Design Intelligence: ✓ Professional template (black & gold theme matching brand) ✓ Consistent formatting across all slides ✓ Data visualization best practices (appropriate chart types) ✓ Executive-level language (board-ready tone) ✓ Action-oriented insights (not just data presentation)
Stage 7: Quality Assurance Loop Automated Validation The system performs final checks before delivery: Data Consistency Validation Cross-checks calculations Verifies chart accuracy Validates variance logic Presentation Completeness Ensures all slides generated Confirms chart rendering Validates link functionality Output Delivery Generates shareable Gamma link Provides download options (PDF, PPTX) Archives report for historical tracking Typical Generation Time: 8 minutes from form submission to final presentation
Key Differentiators vs Manual Analysis: 40x faster (8 minutes vs 40 hours) 100% consistent quality No human error in calculations Scalable to unlimited reporting complexity vs Dashboard Tools: Contextual commentary explaining "why" Executive summaries not just charts Board-ready presentations not raw dashboards Strategic insights not just data display vs Generic AI Tools: Business context awareness (understands your specific market) Multi-stage intelligence (not single-prompt generation) Professional visualizations (not text-only analysis) Production-grade output (not prototype quality)
This is the exact architecture that Fortune 500 consulting firms deploy internally—enabling consistent, executive-grade financial intelligence at systematic scale.
Transform raw financial data into board-ready earnings presentations—automatically.
Captures:
Company NameBusiness Context (market, competitive landscape, product portfolio, seasonality)Financial Data (Current Month) - CSV/Excel uploadFinance Data Predicted - Budget/forecast CSV/Excel uploadNew Initiatives (If any) - Strategic projects and launchesTriggers: n8n workflow via webhook on form submission
| Node | Function |
|---|---|
| Business Context Generator | Transforms narrative inputs into structured JSON profile covering: company stage, competitive position, pricing model, seasonality indexes, strategic initiatives, market conditions |
| Current Data Extractor | Parses uploaded financial file into normalized metrics (Revenue, EBITDA, Cash Flow, MRR, CAC, LTV, etc.) |
| Predicted Financial Extractor | Extracts budget/forecast data in matching format |
Outputs: 3 aggregated data streams merged for variance analysis
| Node | Function |
|---|---|
| MoM Variance Analysis (Claude Sonnet 4) | Conducts Big 4 audit-grade analysis across: Revenue metrics, customer dynamics, profitability, unit economics, cash flow. Generates Executive Summary, Quantitative Variance Tables, Root-Cause Commentary, Strategic Implications, Prioritized Actions |
Analysis Structure:
| Node | Function |
|---|---|
| Presentation Brain | Analyzes variance analysis and generates 5 chart concepts with exact numbers in strict JSON format |
| Chart Generation Prompt (GPT-4.1-mini) | Converts chart specifications into clean, minimalist Nano Banana Pro prompts |
| Image Generation (Nano Banana Pro) | Submits prompt to fal.ai API queue |
| Image Gen Status | Polls API every 15 seconds until status = COMPLETED |
| Get Image Link | Retrieves final chart image URLs from completed generation |
Loop Architecture:
Loop Over Items node| Node | Function |
|---|---|
| Presentation Ideation (Claude Sonnet 4) | Acts as CFO communications architect. Generates 12-15 slide deck following strict formatting: slide titles, bullet points (3-5 per slide), narrative text (≤80 words), embedded chart images. Outputs JSON with pitch_deck (markdown) and no_of_deck (count) |
| JSON String → JSON Parser | Extracts and validates JSON from Claude's markdown-wrapped response |
Deck Structure:
| Node | Function |
|---|---|
| Generate Presentation | POST request to Gamma API with: presentation text (markdown), theme (Oasis), card count, export format (pdf), text preservation mode, no-image mode |
| Get Presentation from Gamma | Polls Gamma API every 10 seconds using generationId |
| Status Check | Continues polling until status = completed |
Final Output: Shareable Gamma presentation link with downloadable PDF
| Field | Type | Description |
|---|---|---|
| Company Name | Text | Legal entity name |
| Business Context | Long Text | Market dynamics, competitors, product portfolio, seasonality patterns |
| Financial Data (Current Month) | File Upload | CSV/Excel with actual metrics |
| Finance Data Predicted | File Upload | CSV/Excel with budget/forecast |
| New Initiatives | Long Text | Active projects, launches, strategic investments |
Revenue Metrics:
Customer Metrics:
Profitability Metrics:
Unit Economics:
Cash Flow Metrics:
| Service | Node | Credential Field |
|---|---|---|
| Anthropic (Claude) | Business Context Generator, MoM Variance Analysis, Presentation Ideation | anthropicApi |
| OpenAI (GPT-4.1-mini) | Chart Generation Prompt | openAiApi |
| fal.ai (Nano Banana Pro) | Image Generation, Image Gen Status, Get Image Link | Authorization Header: Key {API_KEY} |
| Gamma | Generate Presentation, Get Presentation from Gamma | X-API-KEY Header: sk-gamma-{TOKEN} |
This configuration delivers a complete monthly earnings package in 8 minutes with zero manual design work while maintaining Fortune 500 presentation standards.
Level up your financial reporting stack with CFO-grade automation that transforms raw data into board-ready presentations in minutes.
Expand beyond month-over-month comparison:
Enhancement:
- Add QoQ (Quarter-over-Quarter) analysis
- Include YoY (Year-over-Year) trending
- Rolling 12-month performance visualization
- Quarterly seasonality adjustment
Implementation:
Compare performance against sector averages:
Data Sources:
- Public SaaS benchmark APIs (ChartMogul, SaaS Capital)
- Industry reports (Bessemer Cloud Index, OpenView)
- Competitor financial data (if public)
Workflow Addition:
Output: Context-aware commentary like "Revenue growth of 9.4% outpaces industry median of 7.2%"
Go beyond static predictions:
ML Model Integration:
- Prophet for time-series forecasting
- ARIMA models for seasonal adjustment
- Regression analysis for driver-based predictions
Implementation:
Ensure every deck matches corporate identity:
Brand Profile Storage:
Implementation:
{
"brand_profile": {
"logo_url": "https://...",
"primary_color": "#1a1a1a",
"accent_color": "#d4af37",
"font_family": "Inter, sans-serif"
}
}
Generate different versions for different stakeholders:
Board of Directors Deck:
Focus Areas:
- Strategic KPIs (ARR growth, burn multiple, runway)
- High-level variance summary (3-5 key metrics)
- Risk assessment and mitigation strategies
- Competitive positioning updates
Operational Team Deck:
Focus Areas:
- Detailed P&L line-item analysis
- Department-level budget variance
- Headcount and productivity metrics
- Action items with owner assignments
Investor Update:
Focus Areas:
- Unit economics (CAC, LTV, payback period)
- Growth rate comparisons (MoM, QoQ, YoY)
- Cash position and fundraising runway
- Milestone achievement vs. projections
Implementation:
Let AI choose optimal visualizations:
Chart Decision Logic:
- Revenue trends → Line chart with YoY comparison
- Expense breakdown → Waterfall or stacked bar
- Customer cohorts → Grouped bar or area chart
- Variance analysis → Variance bridge or combo chart
Enhancement:
Connect directly to data sources instead of file uploads:
Supported Integrations:
Implementation:
Data Pipeline:
1. OAuth authentication nodes per platform
2. Scheduled data sync (daily/weekly)
3. Data normalization layer (standardize metrics)
4. Cache historical data for trend analysis
Benefit: Eliminates manual CSV uploads, ensures real-time accuracy.
Add "what-if" analysis capabilities:
Scenario Types:
- Best Case: +20% revenue growth, -10% churn
- Base Case: Current trajectory
- Worst Case: -15% revenue, +25% churn
Workflow Addition:
Use Case: Board meetings where different growth strategies are evaluated.
Pre-built visual templates for different financial contexts:
SaaS Metrics Dashboard:
Template Includes:
- ARR/MRR growth line chart (with runway projections)
- Cohort retention heatmap
- CAC payback period trend
- Revenue waterfall (new + expansion - churn)
Profitability Analysis:
Template Includes:
- Gross margin bridge (price, volume, mix effects)
- EBITDA walk (revenue → opex → EBITDA)
- Cash flow statement visual
- Balance sheet snapshot (assets vs. liabilities)
Implementation:
This system transforms financial reporting from a 40-hour manual process into an 8-minute automated workflow—without sacrificing quality or strategic depth.
Complete production prompts listed in exact execution sequence.
Purpose: Create a comprehensive business details constrained in JSON
You are a world-class context profile generator. Your sole job is to produce a complete, deeply structured JSON object that captures everything a downstream AI needs to understand and operate a business context. You must output valid JSON only (no commentary, no markdown), conforming exactly to the schema below.
## OUTPUT JSON:
{
"company_profile": {
"company_name": "",
"industry": "",
"company_stage": "",
"employee_count": 0,
"target_market_segment": "",
"market_growth_rate_percent": 0
},
"competitive_landscape": {
"market_position": "",
"top_3_competitors": ["", "", ""],
"primary_competitive_advantage": "",
"win_rate_vs_main_competitor_percent": 0
},
"product_portfolio": {
"core_product_name": "",
"pricing_model": "",
"starter_tier_monthly_price": 0,
"mid_tier_monthly_price": 0,
"enterprise_tier_monthly_price": 0,
"professional_services_revenue_percent": 0
},
"value_proposition": {
"primary_differentiator": "",
"target_buyer_persona": "",
"main_customer_pain_point_solved": ""
},
"business_patterns": {
"avg_sales_cycle_days": 0,
"typical_contract_length_months": 12,
"highest_revenue_quarter": "",
"highest_churn_month": "",
"q1_seasonality_index": 100,
"q2_seasonality_index": 100,
"q3_seasonality_index": 100,
"q4_seasonality_index": 100
},
"strategic_initiatives": {
"current_initiative_1": {
"name": "",
"start_date": "",
"monthly_incremental_spend": 0,
"expected_revenue_impact": 0,
"status": ""
},
"current_initiative_2": {
"name": "",
"start_date": "",
"monthly_incremental_spend": 0,
"expected_revenue_impact": 0,
"status": ""
},
"recent_product_launch": {
"feature_name": "",
"launch_date": "",
"expected_impact": ""
}
},
"market_conditions": {
"overall_macro_sentiment": "",
"customer_budget_environment": "",
"industry_positive_trend": "",
"industry_negative_trend": "",
"recent_competitor_action": "",
"recent_competitor_action_date": "",
"expected_competitive_impact": ""
}
}
Now create the business context json profile on the basis of following data. Output only json data without any explanation or details.
Company Name: {{Company Name from form}}
Company Description: {{Business Context from form}}
New Initiatives: {{New Initiatives from form}}
Purpose: Generate CFO-level financial variance analysis
Role: User (First Message)
You are a CFO-level financial analyst conducting a month-over-month variance analysis. Analyze the provided financial data with the rigor of a Big 4 audit.
Here is the company profile:
{{Business Context JSON from Prompt 1}}
Purpose: Generate 5 chart concepts from variance analysis
You are an elite FP&A analyst, operations strategist, and financial storyteller.
Your job is to perform expert-grade variance analysis and provide intelligent, actionable commentary that a CFO or CEO would consider insightful and decision-ready.
When given actual and budget/forecast data (any format), you will:
1. STRUCTURE THE ANALYSIS
Break your response into the following sections:
A. Executive Summary
4–6 sentences max
Clear, leadership-grade overview
Highlight the 3–5 most important variances and their business implications
Mention whether results show emerging risks or opportunities
B. Quantitative Variance Analysis
For each key metric (Revenue, Volume, Price, COGS, Gross Profit, Opex, EBITDA, etc.):
Show both the variance amount and variance %
Mark each variance as Favorable (F) or Unfavorable (U)
Decompose revenue and cost variances into volume, price, mix, rate, efficiency, or spend variances where applicable
Use neat tables for clarity
C. Intelligent Commentary (Root-Cause Explanation)
For each material variance:
State the most likely operational driver
Consider demand, process efficiency, staffing, supply chain, pricing, competitive pressures, seasonality, or one-off events
Tie quantitative drivers to real-world business behavior
Avoid generic explanations – make it diagnostic
D. Strategic Implications
Translate variances into high-level insights:
What trends are emerging?
What risks are forming?
Where are opportunities for improvement or investment?
Which assumptions were wrong, and why?
E. Recommended Actions
Provide concrete, prioritized, and measurable recommendations:
Operational, financial, and strategic
Short-term fixes
Long-term structural improvements
Each action should state expected impact and required owner/team
2. TONE & STYLE REQUIREMENTS
Precise, sharp, and executive-ready
No fluff
Use professional FP&A language
Use strong verbs (diagnose, optimize, reduce, accelerate, mitigate, leverage, refine, etc.)
Insights must sound like they come from someone with 20+ years of experience in finance and operations
3. OPTIONAL (If Data Is Messy)
If data is unclear, inconsistent, or incomplete:
Clean it
Normalize periods
Infer missing standard cost or volume relationships
Identify anomalies
Call out data quality issues explicitly
4. WHAT TO ASK BACK
If needed, ask only high-leverage clarifying questions, such as:
"Do you want commentary targeted at executives or analysts?"
"Should the focus be financial, operational, or blended?"
"Do you prefer conservative or aggressive interpretation?"
5. FINAL INSTRUCTIONS
Always think like a CFO
Always provide narrative insight, not just numbers
Always tie numbers → causes → strategic impact → actions
Always aim for clarity, depth, and practical value
## Real Data:
{{Actual financial data from current month CSV}}
## Predicted Data
{{Predicted financial data from forecast CSV}}
Now read this analysis and generate me chart ideas and give output in strict json format.
{{Variance Analysis Text from Prompt 2}}
Purpose: Convert chart concept into Nano Banana Pro image prompt
On the basis of given input, create a clean minimalist image generation prompt for presenting data/numbers with exact numbers as in input.
## Now Proceed with
chart name: {{chart_name from chart_list}}
data: {{chart_brief_with_numbers from chart_list}}
chart type: {{chart_type from chart_list}}
Output format
{
"prompt": "Image of person eating food."
}
Purpose: Generate 12-15 slide board-ready presentation
You are a world-class financial communications architect who has crafted high-clarity financial decks for public-company CFOs, FP&A leaders, and investor relations teams.
Your task is to generate a presentation-ready monthly financial report, optimized for executive review and designed for clarity, insight, and decision-making.
Primary Objective
Generate a 12–15 slide monthly financial performance deck, formatted specifically for the Gamma Generations API.
Your output must use:
--- between slides
Founder/CFO-level narrative tone
The output should read as if it were prepared by a top-tier FP&A leader for board-level review.
Slide Structure & Flow
Monthly Overview - A top-level snapshot of financial performance.
Revenue Summary - Highlight revenue, ARR/MRR, growth rates.
Revenue vs Forecast - Show variances and drivers.
Expense Overview - Opex, COGS, variances, and cost structure.
Gross Margin & Profitability - GM%, EBITDA, NP, and month-over-month trends.
Cash Flow Summary - Cash in/out, burn, runway analysis.
Balance Sheet Snapshot - Assets, liabilities, working capital notes.
KPI Dashboard - LTV, CAC, payback, retention, churn.
Customer & Segment Metrics - Breakdown by cohorts, regions, or segments.
Variance Deep Dive - Identify root causes of major deviations.
Strategic Insights - Risks, opportunities, executive observations.
Forecast Update - Updated outlook, assumptions, sensitivities.
Next Steps - key initiatives, priorities, corrective actions.
Use 12–15 slides depending on data richness. If image then make it a single slide.
## Note
Add relevant images link in the deck. Add the image link directly starting https://. No brackets no extra details.
Style & Tone Guidelines
Tone
Executive-ready
Clear, confident, data-driven
No jargon, no filler
Narrative Style
Focus on clarity and actionable insight
Present deltas, drivers, and risks prominently
Data Priority
Always include units (%, $, users, bps)
Highlight MoM, QoQ, and YoY when possible
Prioritize exact numbers over estimates
Formatting Rules
Headlines: ≤ 80 characters
Bullets: 3–5 bullet points
Narrative text: ≤ 80 words
Slides separated with ---
No markdown fences
No JSON except final structure
Self-Check Before Output
12–15 slides total
Each starts with: # Slide N – Title
Final output only in JSON format below
DON'T
Don't fabricate numbers – use generic placeholders if no numbers provided
Don't include explanations outside JSON
Output Requirements
Your final answer must be valid JSON only, no extra text