5 Lessons for Harnessing AI in Corporate Finance

David Conte, CFO of Databricks, and NEA’s Mark Hawkins discuss how AI is reshaping data, governance, and forecasting for modern finance leaders

As AI evolves from experimental to essential, leaders are learning to navigate a new competitive frontier. Few understand what this means for corporate finance and administration better than Dave Conte, CFO of Databricks. With more than 30 years of experience in multinational public and private technology companies, Dave has helped multiple category-defining companies scale, transform, and go public.

To kick off our AI Stack Series, Dave joined Mark Hawkins, former president & CFO of Salesforce and an NEA Venture Partner, for a conversation about the foundations of AI in corporate finance. Mark has worked in tech for more than 35 years, adapting to major transitions like the rise of the internet, cloud computing, mobile technology, and social media. In this edited transcript of their conversation, Mark opens by suggesting the transition to AI feels different.

Let the data decide: From opinion to single source of truth

Mark Hawkins: Over the course of our careers, data has been super important. But now, with the advent of AI, data is becoming even more critical. How are you thinking about that?

Dave Conte: From the finance point of view, we've always been responsible for providing information to our constituents to enable decision-making. But with today’s technologies, data volumes are growing, as are the number of data sources that you have to wrangle. Importantly, the speed at which you need to deliver answers has accelerated. 

If there's one word I would use to describe our whole product strategy, it's unification. That means unifying massively disparate data sources into a single source of truth, and governing, enabling, and securing access to that data across the enterprise and inside of finance specifically.

MH: When you're making a critical decision, opinions are great, but what you really want is facts. And that means having a single source of data.

DC: You're 100% right. We have a very strong corporate value, and I think it should resonate with every finance leader: Let the data decide.

Govern data access as if it were a bank vault 

MH: Controlling access to that data is like controlling a bank vault. I would love to hear your thoughts on that.

DC: At Databricks, we have something called the Unity Catalog, a dashboard that enables you to basically grant access to various data sources, and even to code. In our financial model, Unity Catalog is probably the single most important bit of technology that we deliver to our customers. We don't charge for it. It's embedded in the platform because it's fundamental. If you don't have governance of your financial data, then you're back to ground zero. 

MH: These are fundamental principles in the world of AI. Not only to get it right so we can use it, but so we can protect it and treasure it too. Any words of wisdom?

DC: You've got to partner with your security professionals. It's really data security that matters. The real IP of the company isn't necessarily the code that the products are written on. It's all of the customer data—its processes, and what you think about go-to-market. The heightened level of access that you need to give people in the company has grown exponentially, which means you have to do a better job protecting the entire data state.

MH: I completely agree. That's a big takeaway that’s relevant to all finance leaders: Really pay attention to who gets access. Stewardship of data isn’t just about making sure it is valuable for AI applications, and to ensure you make quality decisions at speed. It’s about protecting and controlling it. 

AI in corporate finance: From reporting to real-time intelligence 

MH: What has intrigued you about AI and machine-learning use cases in finance, specifically? 

DC: We have what I would consider a fairly complicated financial model that’s based on consumption. You've got 20,000 customers that have a plethora of SKUs and therefore different pricing, and tons of data that originates with the product itself. Having to tether financial and business systems directly to the product is something I haven’t had to do before Databricks, and it’s been a really fun challenge. We built a revenue application, or revenue subledger, that sits on top of a data lake and speaks to the originating source of data. It has AI-enabled logic that understands ASC 606, the revenue standard. So now we've automated ASC 606 through this massively complex workflow using AI techniques. 

In finance, we also partner with our sales strategy and ops teams around target-setting. We're setting quotas for individuals based on the prediction of how much usage will grow inside their territory. It’s very customer-specific work that takes into account their maturity and plenty of other inputs. It's an example where you've got two organizations, one in sales, one in finance, that are partnering on the work using the same data set. If you had to reconcile two disparate data sets, it would slow the business down. But with a unified data set, we can use predictive analytics and machine learning to effectively forecast usage by customer to set quotas for individuals. 

These are examples of applications that are really intelligent, that are built atop critical data, and can accelerate the value you deliver back to the enterprise.

MH: It’s data that is well governed and protected, which becomes a launch pad for some amazing stuff. And so if finance was strategic before, it's even more strategic today, in the age of AI. 

When we think, “Someday, finance is going to be impacted by AI,” that someday is today. You touched on demand forecasting by region, by product, by customer type, and really analyzing your pipe in ways that are a lot more profound. Everyone who's ever done this knows it is hard and tedious. Today, it’s happening automatically thanks to AI. Automated reports, reconciliation, variance analysis, exception routing. That’s all happening today. People are doing anomaly detection today. Instead of doing a sample audit manually, you do a 100% audit, and you do it in about a picosecond.

DC: We've historically been in the business of reporting what happened. Enabling your business to have a better view of what is going to happen, especially in a consumption model, with accuracy and confidence, is a true value differentiator. 

Financial data accuracy and AI are key to investor trust 

MH: Where else is AI becoming mission-critical?

DC: Think about travel and entertainment. We now process 250,000 transactions a month with just five people. We're not talking about chatbots here. We're talking about using technology to audit accuracy and compliance. That’s not to save money. That's to keep people out of trouble. 

MH: That’s another powerful example. And it’s especially important for companies that plan to go public someday to think about these issues. Being public, you’re on a much greater stage, and you need to protect the trust you have with investors. You really want to use all the powers that you have before you to make sure you earn that trust. And it all goes back to data governance, data unification, and data management, allowing you the time to model and have insights. Forecast accuracy is foundational to the value of your company and the optionality that your company has strategically. 

DC: Absolutely. Just yesterday, I was speaking to an investor who said they track all companies’ financial reporting versus their outlook, and tag them as red, yellow, or green. He said Databricks is green. But the whole point of the exercise isn’t to determine “Are you good or bad?” It's to determine “Are you credible and can we trust you with our LPs' money?”

MH: If you're not credible, that money costs more — if you get access to it — and your valuation is less. If you are credible, the risk is different. So this is foundational.

Lean into the future

MH: AI is enabling new business models. What are the implications for finance professionals? 

DC: I was brought up in the era of on-prem software and then SaaS subscriptions. The consumption model, from a forecastability perspective, is harder. But I actually like it a lot because it directly aligns with the value a customer gets. If you're thinking about your pricing models and how you build your infrastructure and go-to-market, don't solve for what's easy within the infrastructure teams. Solve for what correlates to value to the end user. Because if they're getting value, it shows up in your revenue. And if they're not, it's a signal you've got to do something different. 

In the old model of enterprise selling, there’s always that push to get a deal done by the end of the quarter. That’s important. But making sure your customer is consuming the product is more important. The whole mentality of the company has to move in that direction. You want to be an enabler of that. Structure your company to do that, then build all your systems, your infrastructure, to do that really well. Then, if you’ve got product-market fit, you're going to have a good outcome.

MH: Any closing thoughts on what it means to be a corporate finance leader in the age of AI?

DC: What I'm excited about is getting our people out of the data wrangling game and the data reconciliation game so they can become way more productive and help the business advance itself. 

I think where we are and where this whole thing is headed is so exciting. I would just be aggressive with getting that infrastructure for these technologies in place. Invest in that infrastructure right up front. Get that data accessible and ready to go, and then you'll be leading the company with analytics, versus holding it back.

MH: I couldn’t agree more. I’d say don’t be afraid to get on the risk frontier even earlier in your career. Understand it, manage it, and lean into it. Invest in yourself and your development to meet the moment, because the future is going to come at you, and fast.