It’s easy to fall into the trap of thinking that a finance team can accomplish anything with nothing more than an AI subscription these days. That you can point an LLM at the financial data and it will produce forecasts and board narratives on demand.
But you know it’s not that simple. AI generates confident, polished outputs regardless of the data feeding it, so a messy revenue file and a clean one can both produce authoritative-looking analyses.
AI can only drive ROI in finance if the underlying data is unified across systems and standardized for clear, real-time analysis.
To explain what it takes to put that data foundation in place we sat down with Nidhi Shah (Global VP of Finance at Level Access) and Paul Barnhurst (The FP&A Guy) for our webinar, How Data Strategy Drives AI Strategy in Finance, where they laid out a data strategy methodology that maximizes the value and potential of AI.
The Methodology for an AI-Ready Financial Data Strategy
Data chaos isn’t a new problem for finance teams. However, the pressure to get the most out of AI is putting a new spotlight on it. Nidhi Shah is looking at that as a blessing in disguise.
“AI pressure presents a real opportunity for companies to use it as a forcing function to create an investor-grade data infrastructure. I see it as a blessing in disguise. It’s fun to navigate, but I think you have to be very methodical in how you approach it.” — Nidhi Shah, Global VP of Finance, Level Access
To create the investor-grade data foundation for her AI initiatives at Level Access, Nidhi and her team spent about 10-12 weeks working through a data strategy framework.
There were four critical steps you should consider if you want to build the foundation in your org.
Choose and Standardize Your Core Metrics
A workable data strategy starts small: four or five metrics that actually drive board decisions, like ARR, NRR, GRR, LTV, and CAC. The alternative is trying to "boil the ocean," in Nidhi Shah's words, and watching the project collapse under its own scope.
But even a short list only works if everyone agrees on what the numbers mean. At Level Access, ARR meant something different depending on whether a customer came through the product-led motion or the sales-led one. Untangling that distinction, and settling on one definition for the board, took real time on its own.
Get the definition right once, and every AI output built on that number inherits the same clarity instead of scaling any confusion.
Assign an Owner to Each Source System
Each source system needs a permanent owner, someone accountable for keeping it clean long after the initial cleanup project ends.
"Who's the owner? Let's get all the systems right, get all the right people in the room, then we'll do the cleanup effort. I've seen too often where you try to clean up all the data, but you haven't fixed the problem. That's just a band-aid. It always starts with the source." — Paul Barnhurst, Founder, The FP&A Guy
At Level Access, three teams split that responsibility along the systems closest to their work:
- Accounting owns data quality in NetSuite, the company's ERP
- RevOps owns data quality in Salesforce, the CRM
- Finance owns the tech stack turning that clean data into reporting
Nobody owns the whole system, and nobody needs to. Each team is accountable for the piece closest to it, which is what keeps a small gap from becoming next quarter's board-meeting surprise.
Fix the Source Systems
If the systems feeding your standardized metric calculations aren't integrated correctly, the number will be wrong no matter how precisely you define it.
In the past, it may have been enough to get by with “good enough” integrations and data hygiene if you could get reporting out on time. But the bar is much higher for effective finance teams now. Having 80% of your data relatively clean won’t mean much if you don’t tackle the last 20% of the issue.
Level Access dealt with that issue on its ARR calculations. They ran ARR directly out of NetSuite, but the CRM feeding customer and revenue data into it wasn't integrated as tightly as it needed to be. Manual patches on the NetSuite side meant the number was right for reporting, but the team needed to clean up the Salesforce side, too, if Nidhi wanted to make the most of AI tools.
You might see issues like this worsen in a PE-backed growth-through-acquisition strategy. After acquiring a product-led business, Level Access had two options to update its ARR calculations at the source systems:
- Good enough. Fold the new data into its existing sales-led motion with limited scope to get a number in place that they could modify later.
- AI-ready. Rebuild a holistic view of ARR across the two different GTM motions so that all source systems fed into a single view of revenue.
The “good enough” path is always faster, but according to Nidhi, “at some point, if you don't have the right integration across different systems, it's just going to slow you down eventually."
You’ll always feel a pull to tackle other initiatives rather than invest more time in data hygiene. You have to resist that urge if you want an AI-ready data foundation.
Automate the Board Narrative
Automating the board narrative means the tech stack assembles investor-grade, segment-level views from clean data automatically, rather than a team rebuilding that view by hand every time the board asks a new question.
At Level Access, Nidhi credits FinQore and Pigment with establishing their investor-grade data and reporting foundation.
Instead of manually recombining data by industry, go-to-market motion, and customer for each new request, the segmented view already exists and stays current on its own. It's flexible enough to extend into margin expansion analysis and Rule of 40 conversations too, without a separate manual effort to build that view from scratch. As a result, Nidhi and her team can close the books and produce a board narrative in a few days, rather than a few weeks.
The tech stack handles that structured, repeatable output on a predictable schedule. Generating insight from it is a separate layer, with its own tools.
AI Tools Build Upon Your Data Strategy
AI tools multiply the value of a data strategy that's already in place. They rarely substitute for one.
Nidhi says that feeding everything to the LLM of your choice “can get you some quick wins, but the mandate is to build something scalable enough to support the business down the road.” And that requires a more robust data strategy and toolset.
Without those pieces in place, AI can still speed up narrow tasks. Judging whether a number is actually right still takes a person who understands what's behind it.
At Level Access, that judgment call belongs to data analysts embedded in the finance team, distinct from whoever's building the models. When ARR shows something unusual, like a sudden spike in downsell, that person digs into it before the number reaches a slide. Only once it passes what Nidhi Shah calls the "sniff test" does it show up in the board deck.
“Ever since we’ve started using these AI-enabled tech stacks, you probably have less value in having someone on the team who can just build, but not deeply understand the data that’s coming out. It’s pivoting from ‘I need modelers’ to ‘I need thinkers,’ and I’ve seen this shift in the recent hires I’ve made.” — Nidhi Shah, Global VP of Finance, Level Access
The glaring problem with everything we’ve discussed here is that you still need to find time to tackle these important hygiene issues while also keeping the business up and running.
It’s one of the biggest reasons we have a managed services approach to our customer relationships. We work with you to put this data strategy in place and implement the AI-powered tech infrastructure to maintain it so you can focus on the business.
If you want to learn more about what that looks like in practice, download our guide for going from data chaos to clarity (specifically in PE-backed environments).
