We’re going to see more artificial intelligence applications than we have in previous years, in a relative sense. But, when you double click on it, what you find is that it’s still very early days.

What we all need to remember is that AI is just a tool for getting insights out of data. It’s still all about the data, because any insights you’re going to get are only going to be as good as the data they’re pulled from.

Your focus now should be on collecting it, being good stewards of it and figuring out where you might use it to add value and find competitive advantages, keeping in mind that AI is not the only way to get value from data. In many cases, simpler, traditional types of analysis may answer your questions. The easiest thing of all might be just to sell your data, if it’s unique and interesting. There are a lot of opportunities, but the data itself has to be valuable. We have to walk before we can run.

It has to start with IT leadership partnering to get their arms around the business systems topology. In a lot of companies, nobody’s really managing this, and every department is more or less doing their own thing. Or, maybe IT is managing it, but more from an applications road map and IT security standpoint, rather than with a CFO’s eye to getting incremental shareholder value from the company’s data.

The only way you can pull all your data together into one common database where you can do something with it is through cloud architecture. But, even if every department is running their own SaaS applications in the cloud, if no one’s looking at how processes intersect and how to bring the data together, you might as well still be using on-premise systems. Finance and IT should be natural allies in ending so-called shadow IT and owning the whole systems topology.

Once you have a handle on all the systems that are in use, the next thing to look at is how you collect data through automation. Finance has been working on automation for a long time now. Our focus initially was on building a more efficient finance function, and most of our applications were for back-office power users. That journey to efficiency has to continue, not just in finance, but companywide, and with a focus on usability, mass adoption and SaaS as drivers of the kinds of data we need.

Power users work in the system and do reports and analysis, but end users are typically the ones that generate the data. If you have automated systems but people aren’t using them you’re not going to get the inputs and you’re going to have incomplete, unreliable data sets. If you want good data, you have to embrace the fact that end users are now as important as power users, and evaluate your technology choices with that in mind.

The next thing to consider is the data architecture. You can’t do AI without huge amounts of data, because that’s what it learns from. One of the big reasons AI is advancing so rapidly is because of cloud architecture, which is creating these massive, specialized data sets. You have to build an architecture within your company that can handle growing amounts of data, in a structure that allows you to normalize it and run machine learning and artificial intelligence on top of it.

What a lot of companies are dealing with right now are multiple silos of data. They might have dozens of ERP systems and other systems that they’re running transactions through, but there’s no way for those systems to talk to each other and share data. You need cloud systems that can sit on top of those systems and pull the data up into a single repository where it can be normalized and extracted in real time.

If you’re behind on this, and still running manual processes and making decisions on data that’s six to 12 months old, you’re only going to fall further behind with respect to artificial intelligence. It’s not just about updating technology by swapping on-premise systems for cloud systems. The name of the game is harnessing the data from those systems to get actionable insights in real time.

That’s the real goal of digital transformation — not automating manual processes, or changing the way the software is delivered. The real goal is to have data flowing through your organization the way blood flows throughout the human body. Automation and the cloud are what support that, and integration is the final piece. Within your organization, you have to have the right integrations to see all the data across your systems. Then you can really start to do some interesting things with data.

Where AI comes in is when you’re applying algorithms to multiple data sets to synthesize the information, find patterns and deliver actionable insights.

Here’s an example of something we’re working on, and probably a lot of other SaaS finance folks are working on, that could eventually involve AI: Trying to predict renewal revenue.

That involves looking at multiple data sources. First, you would look at your own platform data to see how much your customers are using the system. Is their usage trending up or down? Are they using all the seats they bought? Are they hitting certain success indicators?

You might look at your ticketing system, to see what tickets customers have submitted, whether they’ve been resolved or not, and how quickly. You might look at your sales funnel to see if they’re in a cycle to buy any upgrades or add-ons. You could potentially look at social sentiment data. AI could help find hidden patterns among customers that churn, identify churn indicators and flag at risk customers early, while you can still do something to prevent it.

To do that right now, we have to go and pull data from a bunch of different automated systems. Then we have to normalize the data. We’re just starting to see some machine learning applications that can do this. Putting AI on top of it is the next step.

Eventually, AI will be able to help us uncover more and more detailed and nuanced of patterns and insights we can use to make better decisions. But it’s all going to be built on top of data. That’s why in 2018 every CFO should be thinking about data and partnering with IT leadership to make sure their data is trustworthy and accessible. What kind of data do they have, and what kind of data do they need? What do they have internally? What data do their SaaS providers have that they can leverage? What data can they get from their community? What data can they buy? And then, how do you bring it all together, and what do you apply to it? In some cases, the answer will be AI, but the technology isn’t the thing. The data is.