I’ve ignored AI for the last couple years.

Well, not exactly - I’ve ignored AI for FP&A/Finance during that time.

Too much linkedin-style hype about ‘excel is dead’ or ‘replace your finance team with this 1 gpt’… too much noise, not enough signal.

Meanwhile I’ve tried a ton of different AI tools for nearly every aspect of life. It’s become a bit of a rabbit hole for me but has been fun to try out what’s possible (the thumbnail image on this newsletter was made with Gemini, for example).

But as AI continues to pick up steam and prove it’s not just an over-hyped tool, I’m compelled to step back into the arena of AI for FP&A to see what’s fact or fiction.

My two perspectives are coming together:

Over the next few months, my plan is to revisit aspects of the FP&A OS (people, data/tools, reporting, analysis, forecasting, consulting) but with a lens on how AI enables finance outcomes.

Let’s explore the FP&A OS a bit and talk about what I’m hoping we’ll find…

Before we talk specifically about where we’ll be focusing on the application of AI in FP&A, we need to talk about the problem(s) of AI today.

Everyone's talking about AI in FP&A.

LinkedIn is flooded with posts about ChatGPT writing variance commentary and AI "revolutionizing" forecasting. Vendors are slapping "AI-powered" on every product pitch. And your CFO is probably asking why your team isn't using it yet (or they gave you copoilot and crossed it off their list).

So 99% of those who are engaging are struggling to make any real progress…

They're automating the easy stuff that doesn't matter while ignoring the hard stuff that does. They're generating 100-slide decks in seconds without asking if those slides should exist in the first place. They're letting AI write variance explanations for metrics they don't understand themselves.

My personal favorite I heard recently “let’s run our P&L through AI and see if it can forecast better than the team”. What a waste of time (i’ll explain why in a later newsletter).

The real problem isn't that AI doesn't work in FP&A. It's that it amplifies whatever level you're already at.

If your reporting is a mess, AI will help you create messy reports faster. If you don't know how to do root cause analysis, AI-generated insights will be superficial. If your forecasts are built on shaky assumptions, AI will just give you more confidence in bad numbers.

This is why I'm creating this series. Not to sell you on AI, but to show you exactly where it works, where it doesn't, and what you need to have in place before it's worth your time.

We’ll learn together too - I’m not the expert and won’t pretend to be.

I’ll engage with those who are experts to see what’s real. We’ll interview founders of FP&A tools with big promises. We might even build our own AI things along the way.

With that, let’s jump into where I think we’re headed over the next few months using The FP&A OS as our guide…

The dream for the next few years is that (as a leader/exec) you can hire 1 person that is capable of doing the quality/volume of work of 3 people due to their AI skills. Meaning you can scale your finance output without scaling your headcount.

But the path of doing that is unclear still.

I’d actually argue that there’s more efficiency gains to be made by implementing an FP&A tool than by implementing AI. And when I say AI what I mean is thousands of different tools and capabilities spread all over the internet with different levels of security.

As in, you can get that efficiency gain just by picking a better FP&A software stack (AI or not).

For now, I’m not hiring any different. I’m still looking for people who can do more with less (need to be tech-forward), who can interpret complex topics, then communicate them to key stakeholders.

IMO, the rest will work itself out over time.

But I do think that if you’re pride sits in your VBA/Excel skills, you should be thinking deeply about what the field of finance will look like in 10 years from now…

Here’s the main issue with AI - it’s that most executives think it’s new in finance.

My OGs know that machine learning has been around for decades(?) and we’ve had predictive analytics available for a long time.

So we need to distinguish between what’s old and what’s new to understand where we should focus or scrutinize bold claims.

What’s new in AI is:

The sky is the limit and every month something cooler comes out that you can do with AI. That’s what we’re planning on tapping into.

But, I don’t want to lose sight that we are finance professionals, not software developers. We only build or use what will actually drive value for the business - that’s our lens…

Reporting has layers and spectrums. Here’s a few:

So to make a single statement about reporting feels daunting.

Each report, each request, each data pull needs to be evaluated independently. I have a feeling we’ll be spending a lot of energy here exploring what’s possible.

Reporting strategy in the age of AI will likely come down to more a jigsaw-style strategy than a 1-size-fits-all approach.

We’ll need to talk about:

It’s a lot but it’ll be worth the investment of our time…

If you asked any AI right now “what should I be doing with AI in my corporate finance role?” I can almost guarantee it’ll tell you:

Here’s the problem with that (in my opinion)… What’s our job in finance?

To understand what’s going on in the business, form an opinion about what to do different, and work with the business to drive that change.

I have a fundamental issue with trying to outsource any of those 3 items directly.

It’s the same thing as how a ton of LinkedIn influencers are using AI to build their personal brand. There is nothing personal about having AI make content - it’s not your brand anymore.

This one is more of a first-principles bet I’m willing to make. I think there will be finance professionals who use AI to do their variance analysis - and in 3-5 years I think their career will stall. They will have outsourced the single most important part of the job.

So how can/should we use AI for analysis?

It should be a supplement to your variance analysis, a connector of hard-to-see dots in the business, a personal large-data processor, and a way to skip the reporting/dashboarding entirely.

Remember, one of our main goals is to deliver insights to the business.

Historically that meant:

But what if we could just go from forming a hypothesis to communicating insights to the business? I think we can with AI.

This one will be tricky but will be a big unlock if we can figure it out…

Unfortunately this is where most of the AI-slop is showing up on LinkedIn and YouTube.

Influencers will have you believe that building models from scratch is the pinnacle of finance using AI.

And while building a DCF from a prompt is a cool party trick, it’s not where we’ll see our returns as finance professionals. Sure, it’s cool, good, and even helpful. But it’s not our bottleneck (hit reply and tell me if I’m wrong!).

Our bottleneck is something in this space:

Most enterprise finance teams are forecasting hundreds of thousands of data points (expense categories, geographies, business units, product lines, months, years, etc.). We’re constantly rebuilding our models as the business evolves and enhancing our projection process as we learn more about the business.

We’re on a hamster wheel of forecasting and we need tools to help us keep up.

I think AI can help us…

Personally, I’m going to struggle with how I leverage AI for consulting.

Back to my point about the core aspect of finance - it’s literally our job and I don’t want to outsource my job (just the low-value parts).

I think consulting will come down to productivity tools and hacks. Things that help you stay on top of more volume of stuff going on without missing it.

I’m probably underestimating this one quite a bit, but I just don’t see it yet.

Show and tell time! If you’re knee deep in this AI world or a step ahead already, reply to this email. Let me know what you’ve found or what you’ve built. I’ll be sharing cool things on Twitter and will start doing video interviews with finance professionals who are on the cutting edge. If you have a software or workflow, let’s do some show and tell!

Say hi 👋 on LinkedIn or Twitter