Adopting AI in Finance
Lessons from the Controller who built the OpenAI and Rippling accounting teams
Today’s Sponsor: NetSuite
AI is advancing faster than any previous technology shift, and its ability to impact nearly every business function makes adoption both exciting and challenging.
Get your guide to learn how to become an AI champion and go beyond mastering technical skills by developing a mindset centered on continuous learning and cross-disciplinary collaboration.
Adopting AI in Finance
Here are 5 themes I am pushing to my teams about AI:
Using AI should be in everyone’s OKRs
Hire an AI before you ask to hire a human
Re-evaluate all processes to determine how they should change because of AI advancements
Hiring interviews and performance reviews should incorporate AI literacy/adoption
Re-evaluate your tech stack. Can new AI tools make things more efficient?
Adoption of AI in finance is incredibly weak because finance people are typically scared of change and making mistakes.
BUT…that doesn’t mean you shouldn’t be implementing AI and experimenting with what it can do really well (and often better than humans).
A lot of processes that accountants/FP&A folks were doing just one year ago looks archaic today to those people familiar with AI capabilities. I have seen LOTS of examples of this where the month-end close process should be 2+ days shorter or FP&A teams are 30% overstaffed based on current tools available.
Will AI be perfect? Probably not. But it can be REALLY good in certain areas and then it just needs a human to verify and potentially make a few changes.
The problem is that your finance team doesn’t know what is out there today and they haven’t spent enough time experimenting with AI.
To help finance teams get started, I asked one of the most knowledgeable people about adopting AI in finance to explain how she has adopted AI at her companies.
Sowmya is the former Controller of OpenAI and obviously knows a lot about AI. Keep reading for her tips on adopting AI in finance 👇
*Check out Sowmya’s Newsletter and LinkedIn
Adopting AI in Finance: Practical Insights from OpenAI
As the former Controller at OpenAI, I've had an amazing journey exploring how AI can transform finance and accounting. Through a lot of experimentation, my team learned exactly what works and what doesn't when adopting AI. I want to share some actionable steps, real-world examples, and templates to help fellow CFOs and Controllers get started.
Adopting AI at OpenAI
I joined OpenAI in March 2023, a few months after the launch of ChatGPT. The company was scaling up very quickly, and our finance team was less than 10 people at the time. Within a matter of weeks, we outgrew existing manual processes in spreadsheets and needed to move to scaled automation solutions. The catch? We didn’t have dedicated resources to take on these projects, and our internal engineering resources were limited because they were grappling with scaling challenges of their own with the growth of the platform. I researched some software products but the cost and implementation timelines made them non-starters. We didn’t have months to go-live on these tools. We needed to close the books in less than two weeks for the next month-end close.
We turned to ChatGPT and started with small scale Excel → Python script process changes. Over time, our team became really good at knowing where ChatGPT could speed us up and help us build customized solutions that incorporated our data and business logic to get us the outcome we wanted.
Over the last two years, our team saw real results from AI adoption.
Turned a reporting process that took 5+ days of manual effort each month into an automated dashboard producing real-time data throughout the month by leveraging Python codegen in ChatGPT
Built custom GPT bots to answer questions about Accounts Payable invoice coding, travel and expense policy guidelines and more
Leveraged ChatGPT to streamline SOX documentation for process walkthroughs and narratives
Moved Revenue Accounting from Excel to a fully automated SQL based integration into NetSuite, with accountants being able to self-serve SQL queries using ChatGPT for help
The results of this type of practical adoption of AI are remarkable. We reduced our close and reporting timelines drastically, closing by 5 business days each month. We were also able to operate with a very lean finance team – less than 20% of the team size compared to peer company benchmarking data.
4 Steps to Get Started
There’s a lot of buzz about AI in Finance, and I’m here to tell you to ignore the noise and build your strategy from the ground up. Here’s how you can get started.
1. Democratize AI Tools: Let Your Team Experiment
When we first rolled out ChatGPT, the toughest part wasn’t the technology. It was convincing everyone it would actually help. In early 2023, everyone on the team was super excited about AI on the heels of the ChatGPT launch. But when asked about how they’re using ChatGPT to make their work better, the responses were strikingly similar: this is really cool, but it doesn’t really help automate my work.
The turning point for our own team’s adoption was a team offsite that year where I put together a demo of real life use cases for the team to see. We then hosted a “hackathon” where teams self organized into teams, worked on quick build automations using ChatGPT in small groups of 3 or 4 people. In a matter of two hours, the team had put together a wide range of incredible MVP demos and we were able to move many of these ideas into production within weeks.
Here's what worked for us:
Encourage Low-Stakes Experimentation: Start with small tasks. Our finance team used ChatGPT initially for simple things like summarizing budget variances or drafting journal entries.
Automate Routine Questions: If we heard the same finance question multiple times (think expense policies or deal approvals), we created an internal chatbot. This saved hours of back-and-forth emails.
Position AI as a Partner: Always reinforce that AI tools like ChatGPT are there to make life easier—not replace jobs. My team quickly saw that AI gave them more time for analysis and strategic projects, which boosted morale and adoption.
2. Clean Up Your Finance Data for Automation
Good data is crucial for successful AI projects. We prioritized cleaning and centralizing our data to make automation possible. Here's how:
Centralize Data: We moved away from scattered Excel files toward structured databases (Databricks, Snowflake), accessible via SQL. This single step made automation feasible.
Automate Data Tasks: I personally started generating Python scripts using ChatGPT to tackle repetitive tasks like categorizing monthly cloud expenses. This turned hours of manual work into seconds of automated script execution.
Build Skillsets (Without Stress): Even accountants without coding experience got comfortable creating basic automations with AI's help. We held casual learning sessions where the team practiced prompting ChatGPT for scripts.
3. Use AI-Powered Transactional Software
High-volume, repetitive processes (like Accounts Payable or expense approvals) benefit hugely from AI-enhanced software. Here’s how we did it at OpenAI:
Choose Smart Software: We adopted expense management software with built-in AI. It automated compliance checks, caught errors, and learned continuously. This immediately freed up time for strategic tasks.
Real-Time Compliance: AI tools flagged policy violations instantly, greatly reducing the manual workload and human error.
Encourage Continuous Improvement: AI-driven software gets smarter through feedback. Regular team input significantly improved software accuracy and reduced corrections over time.
4. Get Executive Buy-In and Align with Team Goals
To drive meaningful AI adoption, you absolutely need executive sponsorship and clear alignment with your team's broader objectives. Here's how we approached this:
Secure Clear Leadership Support: Getting our CFO and executive leaders visibly behind AI efforts helped immensely. Their enthusiasm trickled down, making adoption smoother and faster.
Integrate AI into OKRs: Embedding clear AI adoption goals (like "reduce monthly close timelines by 3 days" or "automate responses to inbound queries to the deal desk team") into our quarterly OKRs clarified priorities and provided measurable targets.
Tie to Performance: Linking AI project outcomes directly to performance reviews motivated everyone to fully engage. It clearly communicated the importance of AI initiatives to the entire team.
Your Comprehensive AI Adoption Plan
Here's a straightforward, practical guide your Controller or finance team lead can immediately use to begin implementing AI effectively:
Preparation and Team Engagement
Host team learning sessions to introduce AI tools like ChatGPT.
Identify bottlenecks in the close and reporting process, and write down areas with repetitive questions and tasks.
Encourage a culture of experimenting freely with AI.
Data Infrastructure and Automation
Audit and consolidate finance data into structured databases.
Prioritize data clean-up efforts.
Encourage the use of ChatGPT to create automation scripts.
Build guardrails around testing and results validation before shipping to production.
Transactional AI Software Implementation
Evaluate and test software solutions with built-in AI.
Run pilots in high-volume transactional areas (expense management, AP, Order Management).
Measure clearly defined success criteria (accuracy, speed, compliance).
Scale successful pilots across the finance organization.
Executive Sponsorship and OKR Alignment
Gain clear executive sponsorship early.
Embed specific AI-related goals into team OKRs.
Align AI outcomes with individual and team performance reviews.
Continuous Learning and Improvement
Regularly review AI performance and gather team feedback.
Conduct quarterly AI refresher training and updates.
Promote continuous improvement through active feedback loops.
Adopting AI in finance is a journey. By starting small, staying practical, and involving your team along the way, you’ll quickly see transformative benefits like faster closes, fewer errors, and more time to focus on strategic projects.
Footnotes:
Get your guide to learn how to become an AI champion. Every leader needs to champion AI and push AI adoption across the org
Reply to this email if you need fractional CFO or bookkeeping support
Check out OnlyExperts to find offshore finance resources. They have some amazing talent for 20% the cost of a U.S. hire
Sowmya provided a very well thought-out approach. Her insights are indeed practical and very logical.
Love this! I’m Harrison, an ex fine dining industry line cook. My stack "The Secret Ingredient" adapts hit restaurant recipes (mostly NYC and L.A.) for easy home cooking.
check us out:
https://thesecretingredient.substack.com