In the last couple of weeks, the threat of AI-related job-disruption has suddenly begun to seriously impact the stock market:
On January 30, Anthropic launched Claude Cowork plugins for things like legal work and data analysis. Stocks like Thomson Reuters and LegalZoom plunged.
On February 5, Anthropic dropped my favorite model, Opus 4.6. The S&P software and services index dropped 4.6%.
On February 10, Insurify released a ChatGPT-based AI insurance comparison tool. The S&P insurance index fell 3.9%
Also on February 10, Altruist released an AI-enabled tax-planning tool. LPL and Raymond James lost 8%, Schwab lost 7%, Ameriprise lost 6.2% and more.
To those of us who have been keeping close tabs on the development AI, it’s somewhat curious that the stock market is being impacted so suddenly by this news; it’s not like there was any secret that AI was intelligent enough to handle a lot of white collar work or that companies were working to apply it for that purpose.
The whole thing reminded me of something I stumbled upon a few months ago. As with so many of my AI-related discoveries, I found Shortcut.ai while scrolling Twitter when I probably should’ve been paying attention to something else. I just checked my bookmarks, and I still have the original Tweet:
nico@nicochristie
Introducing Shortcut v0.5 The Superhuman Excel Agent It finishes an institutional-grade, 13-tab FP&A model correctly in minutes - 10x faster than humans It won't solve ALL spreadsheet work (yet) but it turns beginners into pros and pros into masters of the craft Try for free
8:38 PM · Dec 3, 2025 · 968K Views
63 Replies · 104 Reposts · 1.59K Likes
Sounds cool, though I very much do not need an agent that can handle institutional-grade, 13-tab FP&A models. I deal mostly in horrendously constructed single-tab P&Ls. But I do spend a nontrivial amount of time turning those into coherent, well-formatted reports for the purpose of evaluating businesses. If it can do the complex stuff, it sure ought to be able to help there.
The next time I had one of those spreadsheets, I pulled it up in Shortcut.
If you don’t want to expand that to see how bad it is, let me give you some highlights:
The first month is Jan 2024, but there are some expenses one column to the left of that, not associated with any month. I assume these are just startup costs from before she started keeping track of things well (I mean, not well, but at all).
Not a problem with the sheet, but amusing that there is a line item for “Budget Coaching” for which she paid $150/month for several months, only to end up with this horror show of a P&L.
There are no formulas on this anywhere. Every number is just a number. A few of them have dollar signs. Most do not.
Some fees switch rows at some point. All the Amazon fees are just “Amazon Fees (storage/shipping/handling) through Dec 2024 then move to other rows in a different section for 2025 (though at least they’re broken out slightly better).
The rows totaling the values in each section are at the top of the section. And also the sections are just incoherent groupings of expenses. And also some of the totals are wrong.
One of my big advantages in acquiring these brands is that while this would cause a lot of people to run for the hills, it has the opposite effect on me. When a P&L is this bad, I can be sure it’s not a sophisticated fraudster trying to pull one over on me with regard to the state of the business, and since I know this person is not financially savvy, it’s almost a guarantee that I can find quick wins in the business to get margins up.
Anyway, I started with this well-considered, thoughtfully-written prompt:
This is a nightmarishly bad P&L I received. Can you clean it up
Response:
I’ll examine the P&L to understand what needs cleaning up.
Getting sheet summary to understand the P&L structure
Reading the full P&L data to see the issues
I can see the issues - this P&L has:
Inconsistent formatting (some cells have
$, some don’t; some have quotes around values)Data in wrong columns (some data in column B that should be elsewhere)
“Monthly totals:” labels mixed in inappropriately
Missing section headers/organization
Empty rows creating gaps
Inconsistent number formats
Clarifying Questions
1. How would you like the cleaned P&L organized?
<Three choices presented to me, of which I chose “Create a new sheet with clean structure, keep original for reference”>
2. What should happen to the misplaced data in column B (e.g., $131 Bank Charges, $900 Freight)?
<Three choices, I chose “Keep them but flag for manual review”>
3. How should totals be handled?
<Two choices, I chose “Replace hardcoded totals with SUM formulas”>
Sure enough, it gave me a new sheet that was well-formatted and actually used formulas. Next I had it move the Amazon fees all onto one line and make a couple of adjustments to reflect the numbers in the way most relevant for my analysis of the business:
Convert the COGS from cash to accrual
Get rid of all the one-time and other irrelevant costs
Perfectly done on the first go. This one wouldn’t have been hugely time-consuming, but I probably saved an hour or two. I’ll take that! Also Shortcut gives you 30 credits/day for free, which was sufficient for this task, so I didn’t even have to pay for this. Here’s what we got, post glam up.
Nice!
A few months later, I found a listing for a brand that looked promising but that also represented a new low even for me in terms of financial documentation. Didn’t have a P&L at all, just the Amazon transaction report. That’s just a sheet containing a line item for each transaction (basically sale or fee charged) for the entire year.
Could I turn that into a P&L? Pretty much (no COGS on there, but otherwise everything’s in that report). Do I want to do that? No. I am lazy, and that would be a boring several hours.
You can guess how this goes. Gave it to Shortcut, asked it to generate a P&L, made a sandwich, came back to a P&L. Margins ended up being way too low to buy after I found out landed unit costs, but no harm there since I did no work except to command the ghost in the machine.
This is all to say that if you run a small business and need help with spreadsheet stuff, I wholeheartedly endorse Shortcut (and to be clear, they’re not paying me for my endorsement, but if they’re reading this I am willing to offer a pretty cheap newsletter sponsorship).
To circle back to my initial point, we have had a very capable Excel agent (and to be clear, I know there are others like Claude in Excel, though I haven’t had the chance to try them yet) for months. So why the sudden impact? People working in Excel represent a meaningful amount of white collar work in America, after all.
Perhaps it’s that Excel use tends to be a tool used in a business, rather than an entire business on its own, and the releases that caused stock declines are much closer to letting consumers do work on their own that they previously could’ve paid for.
I can certainly see why a Claude Cowork legal plugin would hurt LegalZoom — it’s probably not something anyone’s going to use for complex contracts, but one certainly might turn to AI for simple tasks like LLC formation and simple agreements. Similar with Altruist’s tax planning features. If you’ve got a billion dollars, you’re going to keep employing tax experts, but there are a lot of people out there with a few million whose tax situations could likely already be handled optimally by AI.
The whole thing lays bare an interesting divide in how AI tools are likely to affect the stock market. On the one hand, products like Shortcut may be good for businesses; one can imagine that a powerful Excel agent will save investment banks from having to hire as many people but will probably not cause their clients to stop hiring them.
On the other hand, it’s not tough to imagine those products expanding their surface area until they do become full business replacements. A bookkeeper might be able to make great use of Shortcut to save time and staffing costs, but before long might find that Shortcut has added enough features to become a direct competitor.
If you’re in a consumer-facing business, and your entire business can be done on a computer, I suspect you’re going to be in for a rough time in the next few years. When everyone has a digital assistant who is smarter, faster and more capable than any human they can hire, it certainly doesn’t bode well for the humans looking to get hired.
One of the reasons I’m writing this Substack is to highlight what that shift looks like in my business. I’ve got Claude doing my books and Nano Banana Pro handling product photos. Most of my contracts are being written and reviewed by AI (some lawyers may be aghast, but the reality is for my deals it doesn’t make sense to spend thousands on a human lawyer for a number of reasons).
AI’s not doing my taxes yet, but I did throw last year’s return at Claude and got several suggestions of things to look at this year that my CPA agreed were valid. By next year, if my CPA doesn’t have a plan for adopting AI and bringing my bill way down as a result, I’ll be looking for a new one.
It’s a great time to be a business owner or a consumer who uses white-collar professional services, but not so much if you’re on the other side of it. The woman who paid $150/month for budget coaching and ended up with a travesty of a P&L can now spend $20/month for excellent budget coaching, a high-quality P&L and a fantastic business advisor to boot. The person selling that budget coaching is going to have a hard time getting clients (though from what I’ve seen, that’s probably how it should be).