Still stuck in spreadsheets? you’re not alone

My background is in finance.

I started in investment banking which may have just been the pinnacle of my exposure to how businesses rely on strategic advisors to craft stories with their data while observing the external forces (“comps”) that ultimately determine whether a market dislocation exists.

This led to private equity where that exposure turned into operational discipline. We built capital intensive companies from the ground up. I spent a good chunk of my time serving as air support FP&A for wholesale data center business. At the time, our capacity was getting rapidly consumed by the hyperscalers leading to pear shape economics for once steady industry (and now I may have been wishing to stick in that space).

From there and post MBA, I took those learnings to run and ultimately acquiring e-commerce businesses on the merits of knowing the CAC to LTV unit economics.

And since, I’ve served as COO and CFO of data infrastructure and application businesses.

The common application among these is that our finance source of truth always remained the spreadsheet.

To those in finance, the spreadsheet is the perfect software.

It’s a language that finance can speak.

Blue cells for hardcodes, green for reference cells, a tab for sensitivities.

Yet for what we gained in simplicity and versatility came the inevitable drawbacks.

From a lack of connectivity to external sources, proneness to manual errors, formula easter eggs, to sprawling tabs.

And while many attempts were made to displace the spreadsheet (shoutout to Airtable), nothing quote could jockey this imperfect perfect tool away.

That’s until claude code ushered in the AI building era.

Finance people inherently are process driven, and at least for me, claude code becomes the conduit to exercise that inhibition.

Not surprisingly the processes that revolve around the manual spreadsheet came first and the Frankenstein, future looking genius of the spreadsheet product is that it's already well structured for how AI consumes information.

The growth models and manual customer tracking, from cohorts to billing were the first targets I was able to have these models consume.

And paired with connections to the upstream source systems like Stripe and QBO, the once almost impossible task of keeping multiple disconnected data sources in sync started to break down.

Don’t get me wrong, this is not just some point and click integrate and prompt situation. There’s a reason why many systems are not compatible, they lack a data governance layer that allows them to speak seamlessly.

But once again, that is where AI can as a foundation as long as you yourself have a good idea of how these sources should talk to each other.

Finance and operations are often the forgotten bits when it comes to where to direct AI first since there’s no perceived ‘growth’ from automating these processes that were once defined by manual reconciliation.

But the physical and mental time that this frees up for the right operations and finance minds leads to more time for unblocking the most critical business scaling issues.

To close, I wouldn’t write the spreadsheet completely out. I still use spreadsheets all the time to consume information as its the love language I know best. What I don’t spend time doing though is constantly worrying about the quality of the data, and whether the other critical systems that I use around the spreadsheet will suddenly break leaving an even larger operational hole than where we started.