This is really amazing, and it reminds me of an Excel modification I did 24 years ago for an animation studio. They had around 3,000 digital artists and then about 200 support staff, and around 8 different productions moving through the studio at the same time. Each production was at some different stage of completion. The studio's production management staff spent a huge amount of time negotiating what staff was assigned to what productions, and the reasonings were all hazy and non-fact based. So I wrote what people would now call a deep learning system that mined all the accounting and production tracking records the studio had, and created probability distributions and confidence intervals for every individual and every task, and every combination of people and their combined tasks, and that all composited together into a giant predictive engine of real time whatif scenarios for the entire studio in different configurations.

It was used for a series of feature films you all probably saw: the Chronicles of Narnia series, the Scooby Doo live action CG films, the Garfield films with the CG Garfield, and Life of PI.

People both loved and hated it. They loved the insights into the studio's capabilities, and they hated thinking through how the thing operated, and they hated the thought process of constructing a new whatif scenario, because they had to THINK.

Wow, this is incredibly cool. I can't believe this stuff hasn't been ported into Excel.

First thing that came to mind is: can this model correlation between each cell, or does it assume that all cells are independent?

For example, in this https://www.getguesstimate.com/models/2205 hours of sleep per night is perhaps correlated with hours per weekday commuting. You might not have the data, but if you do, it could improve the probability distribution in the "weekday hours occupied" cell

As a number of the comments have suggested there are a number of Monte Carlo methods (including this addin) but I was hoping that someone would make one which calculated based on random variables and output a resultant random variable, over brute force monte carlo. Monte Carlo is also somewhat slow, but doing the maths analytically should be fast.

Off-topic: managing @Risk licenses in a fully on-prem, no-internet environment has been such a pain in the neck.

I would argue using Claude in replacing a spreadsheet is potentially more of a tech debt than the spreadsheet itself. Now if you would use Claude to build a point application that performs the same standard analysis it might be better, but only marginally so.

I love Claude and use it for many things, including analyzing and manipulating Excel files. But I’m not convinced that replacing spreadsheets with Claude is the right approach.

Over the years, we’ve seen many platforms promise to replace spreadsheets for a particular use case or industry. Businesses adopt them and push users away from Excel, but in many cases, those same users eventually find their way back to Excel to do the exact same work.

My concern is that we may end up creating another dual-system maintenance problem, where some people use AI and others continue using Excel for the same business process. Over time, the two versions inevitably diverge, potentially creating more problems than they solve.

Excel is much stickier than most people realize. I think the more interesting opportunity is figuring out how AI and Excel can work together, rather than trying to replace one with the other.

Recent grads at all levels always suffer from this problem. The issue is they’ve not needed to use it with enough frequency in their academic lives to build the knowledge or muscle memory. Things have improved a lot in past 10 years or so, but it’s even true with many Finance and Accounting grads (that’s who I hire). Many of them just have never seen real world business data and working with it so different than the contrived information they had seen in coursework. That said give them $15 to spend on Udemy. Then, give them a few assignments, with real world business data. And most people pick it up pretty quickly.

A PhD is not for teaching about specific software tools, and hopefully even less so about legacy software from the 80's.

If you want people to be able to use Excel or service a mechanical typewriter, you need to spend extra effort to teach them.

I'm one of those. Excel's ease of use isn't relevant if you already know how to use Python / Jupyter / DuckDB / whatever.

The mental model of describing a series of reproducible transformations on normalized tables is more intuitive to me.

Excel lets you do too many things. As a result, I mostly see spreadsheets that aren't much more machine-parsanle than a word document.

We had to write a score-table at a party lately, and a while ago, and I managed to get it working with DataSpell faster, than it took us to figure out how Google Sheets stupid date-parsing works.

By the time it takes to make a custom filtered table non extensible UI for an api, the excel guy is finished three times over

I recall excel being decent once upon a time, back in the early 2000's, before the ribbon interface was added.