The last finance function
May 2026 ended finance as a human-only job. Jamie Dimon onstage with Anthropic. 10 finance agents shipped in 30 days. Hub Group's $77M error. 38,579 AI layoffs in a single month.
May 2026 will go down as the month finance stopped being a human-only job.
Jamie Dimon built a Treasury dashboard on a Saturday using Claude. Twenty minutes. Solo.
Hub Group fired their CFO over a $77M accounting error.
38,579 AI-cited layoffs in one month. The highest ever recorded.
Anthropic shipped 10 ready-to-run finance agents that touch the close, the GL, KYC, and the board pack.
87% of CFOs say AI is critical. Only 21% see real value yet.
That gap is your next 12 months. Here is what just happened. What it means for your seat. And what to do before quarter-end.
Things are getting serious.
Let’s dive in.
Jamie Dimon built a Treasury dashboard on a Saturday. By himself.
On May 5, Anthropic launched Claude for Financial Services.
The CEO of JPMorgan was on the stage.
Dimon described what he did the previous Saturday. He opened Claude Code. He asked about asset swaps, Treasury bid-ask spreads, investment grade credit. Twenty minutes later he had a dashboard.
“With all the backup, all the research, and it was very accurate.”
The CEO of America’s largest bank did this on a weekend. With no team. That moment ended the AI is too complex for finance argument.
In the same launch, Anthropic shipped 10 ready-to-run agents inside Claude Cowork and Claude Code:
GL Reconciler. Ties beginning balances to activity to ending balance. Surfaces unexplained deltas instead of plugging them.
Month-End Closer. Builds roll-forwards for inventory, accruals, debt. Flags unresolved items for controller review.
Statement Auditor. Tests financial statements for internal consistency.
KYC Screener. Runs counterparty due diligence at scale.
Valuation Reviewer. Stress-tests assumptions in incoming models.
Earnings Reviewer. Cross-references earnings releases against actuals.
Model Builder. Builds driver-based forecasts from scratch.
Pitch Builder. Drafts investor and board materials.
Meeting Preparer. Builds briefings from CRM and email context.
Market Researcher. Pulls live data from Moody’s, S&P, PitchBook, MSCI.
Every one of these is a workflow your team runs today.
Every one installs in one click.
Goldman’s CIO Marco Argenti said it cleanly:
“This is the first time that instead of buying infrastructure, you can actually buy intelligence.”
The CFOs who install these agents this quarter will run a different finance function by Q4.
Hub Group fired their CFO over $77M. An agent would have flagged it
Hub Group disclosed a $77M understatement of purchased transportation costs and accounts payable across three quarters of 2025.
The stock dropped 27% in one day. Roughly $800M of market cap vanished.
On May 28, CFO Kevin Beth and COO Brian Meents were out. The board hired an interim CFO at $125K a month as a contractor. Nasdaq sent delinquency notices for the 2025 year-end and Q1 2026 filings. Hagens Berman expanded the investor lawsuit on June 2.
That is the most expensive AP coding error of the year.
Here is what should stop every controller reading this. NetSuite, KPMG, Workday, and Ramp all shipped agents in May that catch exactly this kind of mistake before it hits the books.
NetSuite 2026.1 released Intelligent Close Manager.
KPMG and Workday launched the Ignite Financial Close Companion on April 22, built on Gemini.
Ramp Stack shipped June 3 and is benchmarking 4 to 11 percentage points above ChatGPT, Claude, and Gemini on 200+ accounting tasks.
Anthropic’s GL Reconciler does this work natively inside Claude Cowork.
Each one runs continuous validation against the GL. Flags exceptions in plain English. Refuses to post entries that do not balance.
Tyler Otto, who runs an outsourced accounting firm using Ramp Stack, reports clients seeing a 50% reduction in month-end close time.
The tools that would have caught Hub Group’s error before it became a market event are now sitting in your software vendors’ product menus.
The CFOs who deploy them are buying career insurance. The ones who do not are betting they will not be the next Kevin Beth.
38,579 AI-cited layoffs in May. The highest single month ever.
Per Challenger, Gray & Christmas, AI accounted for 40% of all U.S. job cuts in May 2026. A record 38,579.
The third straight month AI was the top cited reason.
The 2026 year-to-date total already exceeds all of 2025.
CRO Andy Challenger:
“AI is now the leading reason companies give for cutting jobs.”
Block cut from 10,000 to under 6,000 in February. Jack Dorsey cited AI directly.
Coinbase cut 14% in May, about 700 roles.
Citi’s Mark Mason said the bank’s headcount will drop by 20,000 from automation.
Standard Chartered announced 8,000 support role cuts over four years.
Innovaccer cut 850 in an “AI-native” pivot.
Now Cloudflare fired over 1,000+ employees.
Inside the finance function specifically, the picture is sharper. 93% of finance leaders say AI is reducing headcount (Personiv 2026). The average company now runs with 17 open accounting roles, up from 5 in 2025 and 2 in 2024.
Then Microsoft AI’s CEO Mustafa Suleyman went on the record with the Financial Times:
“Most of those tasks” in accounting, law, project management, and marketing “will be fully automated by an AI within the next 12 to 18 months.”
Dario Amodei tried to soften this onstage with Dimon. He reached for Jevons Paradox.
“If you automate 90% of the job, then everyone does the 10% of the job. And the 10% expands.”
Pick the one you believe. Plan for both.
Social Signal
Here are the social media posts to read this Sunday.
The Bottom Line
The CFOs who survive the next two years are not the ones with the most data.
They are the ones who pick a workflow this month. Ship an agent this quarter. Have a 24-month headcount plan ready by Q4.
May 2026 was the proof.
Dimon did it himself. Anthropic shipped the tools. Hub Group showed the cost of not having them. 38,579 people lost their jobs to AI in one month.
The last finance function as we knew it is gone.
The next 90 days are not about whether AI changes finance.
They are about whether you change finance with AI, or whether someone else does it for you.
What we are building at finstory
Real reports from real data, for the first group of beta customers who want in.
If you have ever spotted a contradiction in a board document after it shipped, you are exactly who we built this for.
Best regards,
Team at finstory
AI-native platform that turns financial data into board-ready stories.
Find us on LinkedIn | finstory.ai










