A SaaStr AI + Annual Summit deep dive into how OpenAI, Rippling, SnapLogic, and Gorgias are automating finance operations — and the critical mistakes even AI-first companies are making
The Panel: Moderated by Lloyed Lobo – Co-founder of Boast.AI and author of “From Grassroots to Greatness”
- Sowmya Ranganathan – Ex-Controller at OpenAI and Rippling
- Ahsan Malik – CFO at SnapLogic, former VP Finance at BlueJeans
- Kunal Agarwal – CFO at Gorgias (customer support software for e-commerce), former VP Finance at Navan
Quotable Moments
Sowmya Ranganathan (Ex-Controller, OpenAI): “99% of GAAP revenue is going touchless from Stripe all the way to NetSuite. Revenue close basically happens real time.”
Ahsan Malik (SnapLogic): “We ended up cutting essentially a day and a half out of close with AI and more importantly finding revenue that was essentially leakage — things that were entitled that we should have been billing for.”
Kunal Agarwal (Gorgias): “I aspire to be Switzerland, so I try to be kind of a neutral party… I view a lot of my role as the chief accountability officer.”
What These Finance Leaders Actually Built
OpenAI: From 10 to 45 People, Not 300
When Sowmya joined OpenAI in March 2023 (the month ChatGPT Plus launched), the finance team was 10 people. By March 2025, they had grown to ~45 total (30 accounting, 15 finance). But comparable companies their size typically run 200-300 person finance teams.
The automation wins: They automated the hardest problems first:
- Revenue automation: 99% touchless from Stripe to NetSuite
- GPU cost reporting: From 15 days to real-time dashboards
- The Python solution: Teaching CPAs to code with ChatGPT assistance
The key detail: Their Azure GPU reports went from manageable spreadsheets to 9 million rows per month post-ChatGPT launch. Excel literally couldn’t handle it (1M row limit). The solution? ChatGPT helped them write Python scripts that processed in 10 seconds what previously took 10-15 days.
SnapLogic: Finding Hidden Revenue Through AI Agents
As a $100M+ ARR business, SnapLogic runs lean: 4 people in finance, 8 in accounting. But they deployed their first live AI agent internally before selling it externally.
The breakthrough use case: Order form reconciliation
- The problem: Unstructured data from Salesforce, PDFs, DocuSign, customer documents
- The solution: AI agent that cut 1.5 days from close AND found revenue leakage
- The expansion: Legal contract analysis for termination clauses
Key insight: Their CTO nailed it: “The use cases are in the people and the processes.” You can’t buy AI solutions off the shelf — they emerge from your specific pain points.
Gorgias: The Data-First Finance Strategy
Kunal’s approach at Gorgias (customer support software for e-commerce) is different: his finance org includes 6 FP&A, 8 accounting, AND 16 data analytics/engineering people.
The controversial take: “Data by itself is kind of useless. You need to be able to wrap it with a story and a point of view around what that means.”
Their AI wins:
- Predictive customer behavior modeling for usage-based pricing
- Churn risk scoring for customer success teams
- Inbound lead scoring with market data enrichment
- Semantic layer database that answers questions in plain English
The 5 Critical Mistakes Each Speaker Made
Sowmya Ranganathan (OpenAI) – The Automation Evangelist’s Blind Spots
- Overselling the “teach CPAs Python” narrative – This works at OpenAI with unlimited talent access, but telling Series A CFOs to turn accountants into programmers is tone-deaf to resource constraints
- Ignoring the compliance elephant – Zero discussion of SOX controls, audit trails, or regulatory requirements when automating revenue recognition
- Dismissing role elimination too casually – “We just didn’t hire them to begin with” doesn’t acknowledge the human cost of automation for existing teams
- No mention of change management – How do you actually get a finance team comfortable with AI when “finance and accountants didn’t choose this lifestyle to take risk”?
- Oversimplifying the technical requirements – Making it sound like anyone can replicate OpenAI’s data infrastructure with just ChatGPT is misleading
Ahsan Malik (SnapLogic) – The Product Guy Playing CFO
- Conflating product capabilities with finance expertise – Leading with “we’re an AI agent company” instead of finance credibility undermines trust
- Risk framework too simplistic – “Risk, effort, business value” matrix sounds good but lacks specific finance risk categories (regulatory, audit, data integrity)
- Underestimating the people challenges – Acknowledged the center of excellence need but didn’t provide concrete change management tactics
- Missing the integration complexity – Made agent deployment sound easy without discussing data mapping, system connections, or testing protocols
- No mention of customer data sensitivity – For a company selling to enterprise clients, the discussion of proprietary data protection came too late in the conversation
Kunal (Gorgias) – The Switzerland Strategist’s Execution Gaps
- Overbuilding the data team – 16 people in data/analytics for a usage-based SaaS company suggests gold-plating instead of pragmatic automation
- Philosophical over practical – Spent too much time on “data storytelling” theory instead of concrete automation wins
- Semantic layer oversell – The “talk to database in English” demo sounds impressive but lacks discussion of accuracy, limitations, or edge cases
- Avoiding the hard AI and automation questions – Admitted they’re “pretty early on” but then gave advice to Series A companies
- Role clarity confusion – Managing three different functions (finance, accounting, data) creates accountability diffusion, not the “Switzerland” neutrality he claims
The Real Lessons for B2B Companies
Start with your biggest manual pain point that has clear right/wrong answers:
- OpenAI: GPU cost allocation (9M rows → 10 seconds)
- SnapLogic: Order form reconciliation (1.5 days saved + found revenue)
- Gorgias: Customer behavior prediction (better forecasting)
The three-layer automation strategy:
- Data layer: Clean, accessible, trustworthy foundation
- Process layer: AI-assisted analysis and exception handling
- Decision layer: Human oversight with AI recommendations
The hiring evolution (not elimination):
- Fewer junior analysts doing manual work
- More senior people doing strategic analysis
- New hybrid roles: finance professionals who can work with AI tools
What to Avoid (The Mistakes They Made)
Don’t start with the technology – Start with the process pain, then find the right AI tool
Don’t automate without governance – Every speaker underemphasized compliance, audit trails, and risk management
Don’t oversell the simplicity – “Just use ChatGPT” isn’t a strategy for enterprise finance operations
Don’t ignore change management – The people challenge is harder than the technical challenge
Don’t conflate efficiency with effectiveness – Faster closes are great, but strategic insight creation is the real CFO value
The Tactical Playbook: What to Do Monday Morning
For Series A CFOs
Month 1: Assessment
- Audit your close process: What takes longer than 2 days?
- Map your data sources: What comes from where?
- Identify your manual reconciliation nightmare
Month 2: Foundation
- Invest in data cleanliness before AI tools
- Set up proper access controls for AI platforms (ChatGPT Enterprise, not free accounts)
- Define your risk tolerance matrix
Month 3: First Automation
- Pick ONE process with clear success metrics
- Start with 80% automation, 20% human review
- Document everything for audit purposes
For Growth-Stage CFOs
The 70/20/10 rule:
- 70% process automation (reconciliations, data processing)
- 20% analysis augmentation (forecasting, variance analysis)
- 10% strategic experimentation (predictive modeling)
Build vs. buy decision framework:
- Buy: Standard processes (expense coding, bank reconciliation)
- Build: Company-specific logic (revenue recognition, cost allocation)
- Partner: Complex analysis (churn prediction, usage forecasting)
The Uncomfortable Truth About Finance AI
The biggest challenge: Finance teams are culturally risk-averse, but AI requires experimentation.
The successful deployments they described all had one thing in common: Finance leaders who were comfortable with imperfection while maintaining accuracy standards.
As Ahsan noted: “Finance and accountants were not… we didn’t choose this lifestyle to take risk.” But the companies winning with AI in finance are the ones whose CFOs learned to take calculated risks on process innovation while maintaining zero tolerance for accuracy errors.
The meta-lesson: The CFO role isn’t dying — it’s splitting into two tracks:
- Traditional CFOs who focus on accuracy, compliance, and stakeholder communication
- Automation CFOs who build AI-first finance operations while maintaining traditional standards
The winners will be the ones who can do both.
