Finance prepares for an AI future

As new AI chat tools continue to generate waves of hype, worry, and wonder, how’s this for a prompt to try out on ChatGPT: How will finance professionals really be using AI tools like ChatGPT in the next few years?

Sorting out the potential from the pitfalls is going to take time, especially as the tech improves and evolves. We recently spoke with Glenn Hopper, director at Eventus Advisory Group and author of Deep Finance: Corporate Finance in the Information Age, about how finance departments might use AI tech, and what finance professionals will need to consider to prepare themselves for the possible risks and ROIs of AI in the workplace.

This interview has been lightly edited for length and clarity.

What is the current state of AI and chat-enabled AI in finance as you see it?

Today, for 99% of the people who would be using it, it’s a novelty. It has some use that we’ve seen, and people every day are coming up with new uses….But as far as putting production company data into ChatGPT…the proprietary information that you put into GPT is actually, in this public release right now, used to further train the model.

So by your prompts you can be accidentally leaking proprietary company information. I would strongly advise against anyone putting actual company information that they don't want shared in there. It's just the nature of what it is right now, a further refinement in training for the machine learning algorithm that runs it.

In a future in which you can protect your proprietary data, what do you see as potential uses for finance?

The Holy Grail out there is when you can interact with a machine the way you would a junior analyst. Today, we have dashboards that are great. They give you at a glance a look at all of your KPIs, all the metrics you want to track. You open up a web page, boom, they're all right there, it's fantastic. Very high level. Some dashboards, you can drill in and get a little bit more detail.

But if you want to really dive in and figure out…more detail on it, you've got to get a report request to the finance team that says what it is. If you have a fully integrated chatbot that is trained not just on the chat feature, but on your company’s data, on your industry, what you would do in a further level of refinement, this could be voice-operated assistant that you can ask questions, just like you would have an analyst, and get this information in pretty near real time. That’s going to be amazing.

I don’t have a crystal ball, but I would be shocked if within the next 12 to 24 months, there’s not an off-the-shelf software tool that integrates with your company’s financial data, has things in place that keep your data safe, and…is able to interact in natural language and provide this level of detail.

Just like Google Maps means none of us can get anywhere without an app anymore, is there a risk that either institutional knowledge or creativity would get lost to a tool like this?

I think so. The more you automate, the easier it is to say, “Oh, well, that’s another area that we don’t need humans to do.” If you take the most replicable, mindless tasks and automate those, there’s a thought that, “Now we can reduce headcount in finance and accounting.” What I always argue for is that we can do away with the mindless jobs eventually, but there still needs to be a human in the mix to check the work product.

I really push for thinking of spending on technology not as a cost to the business, but as an investment because the more you can automate this low-hanging fruit, this mindless work, the more you can invest in employees who are doing more mindful work. I see the ideal partnership not being human versus machine, but human and machine using the machine as a tool, just like you would excel or a financial calculator to do your job better and to take away those mindless tasks and what you focus on what matters.

What should finance professionals be doing to prepare for a world in which this kind of technology is part of the finance function?

I’m seeing a lot of degrees now that are combined with analytics. When I say finance degrees or business degrees being combined with analytics, well, what does that have to do with AI? But what analytics is and what the subset of AI that is machine learning that chatGPT is a part of, those are all just statistical models. If you have an understanding of the foundation of what this flavor of AI is doing, then you understand how it works and what to look out for more, and when to use it and when not to.—DA