Derek Mills

Anthropic Is Now Inside Your Accounting Software Whether You Agreed to It or Not

March 26, 2026

On February 24, 2026, Intuit and Anthropic announced what they called a "multi-year, game-changing partnership." The press release used the word "groundbreaking" before most people had finished their morning coffee. I read it three times, not because the language was complicated, but because I kept waiting for the part that explained what users actually agreed to.

I'm still waiting.

Here's what happened in plain terms: Intuit's financial intelligence will be surfaced directly inside Anthropic products through MCP integrations with TurboTax, Credit Karma, QuickBooks, and Mailchimp. Run that back. The accounting software that holds your revenue, your payroll, your vendor payments, your tax records - that data now has a direct pipeline into Claude's environment. Intuit has approximately 100 million customers worldwide. One hundred million people. Most of them small business owners who just want to reconcile their bank accounts and go home.

I showed this to Tory before writing this piece. He read it, nodded slowly, and said something like "that's actually great for productivity" with that relentless optimism he carries around like a life raft. I don't know how he does it. The man's car got repossessed last Tuesday and he was still talking about positive disruption at lunch.

What "With Customers' Permission" Actually Means

Intuit's press release and their AI principles page both repeat the same reassuring phrase: "Customer data is not shared with our partners to train their models, and our agreements explicitly prohibit the use of customer data for model training." Fine. Good, even. But that's a very specific promise about a very specific concern, and it leaves a lot of other territory unaddressed.

The phrase that bothers me more is "with customers' permission." It shows up in every version of every press release. "With customers' permission, Intuit applies user data to power high-quality experiences within Anthropic's environment, backed by industry-leading technical and organizational safeguards." That sounds clean. The problem is what Intuit's own privacy policy says about how that permission actually works in practice.

Here's the relevant language from their privacy statement: "Your use of our Service after a notice of material change or posting of an updated Privacy Statement shall constitute your consent to all changes." Logging in and doing your bookkeeping counts as consent. That's the standard SaaS move and I'm not pretending it's unique to Intuit - everyone does this. But it means the phrase "with customers' permission" is doing a lot of heavy lifting on a very loose definition of permission.

And then there's what's already happening in QuickBooks right now, before the full Anthropic rollout even begins. Effective May 8, 2026, the AI-powered bank feed experience will be the standard for all QuickBooks Online and QuickBooks Online Accountant users, and the temporary opt-out option will no longer be available. The opt-out is going away. That's not speculation. That's Intuit's official announcement. You can currently say no. You will soon not be able to.

The QuickBooks community forums have been running hot about this for months. Real users, actual bookkeepers, people who have been using the software for years. One of them put it bluntly: "Please make these 'opt-in', so that regular users are not bothered by this AI nonsense. I'm paying to have a functioning accounting package, don't need any more toys that just interfere with my productivity. If someone wants them: great, let them opt-in." That's not a fringe opinion. Thread after thread, same sentiment.

The Scale Makes This Different

I want to be honest about what Anthropic and Intuit are actually building here, because some of it is genuinely impressive. For example, a regional restaurant group with 15 locations will be able to prompt Intuit Intelligence to orchestrate and deploy Claude within Intuit Enterprise Suite - combining third-party data from their sales and inventory apps with Intuit data such as expenses like food costs, payroll, and workforce hours to automatically highlight margin variances and identify underperforming locations. That's a real use case. That would save a real operations person a real amount of time.

The accounting profession is under genuine strain. Globally, AI adoption in accounting firms surged from 9% in 2024 to 41% in 2025, according to the Wolters Kluwer Future Ready Accountant report. According to Karbon's 2025 State of AI in Accounting report, firms that invest in AI training are unlocking an additional seven weeks of capacity per employee per year. Those are real numbers. If Anthropic inside QuickBooks delivers even a fraction of that value to a small business owner who can't afford an accountant, that matters.

I'm not arguing the technology is bad. I'm arguing the rollout model is broken.

There's a version of this story where Intuit emails every QuickBooks subscriber, explains exactly what the Anthropic integration does, describes precisely which data flows where, and gives users a real on/off switch with no expiration date. That version would face resistance. Some people would opt out. Intuit's data pool for training the experience would be smaller. But it would be honest. That's not the version they chose.

Goldman Sachs, to their credit, is handling their version of this differently. The bank has been working with embedded Anthropic engineers to co-develop autonomous agents in at least two specific areas: accounting for trades and transactions, and client vetting and onboarding. That's enterprise software with enterprise procurement - meaning someone at Goldman signed off on specific contractual terms, with lawyers in the room, before Anthropic got anywhere near their financial data. The 100 million QuickBooks users didn't get that meeting.

This Is a Pattern, Not an Accident

The way this works is that your accounting software starts offering AI suggestions. Nice ones. Genuinely useful ones, sometimes. You use them. You benefit. Then the opt-out window quietly closes - the temporary opt-out option will no longer be available - and the AI-powered experience becomes the only experience. And now here comes Anthropic, plugged in through MCP, with Claude users able to access Intuit's money, tax, and accounting experiences within Cowork, Claude for Enterprise, and Claude.ai.

This is exactly what I was trying to explain to Stephanie last week. She looked up from her iPad and said something like "but the personalization is so much better now" which is true, and also exactly the point. The personalization is better because they're using your data. Stephanie grew up in a house with a tennis court, so the idea of a company having leverage over you through a software subscription lands differently for her. But for a freelance accountant who runs their entire business through QuickBooks? This is their one critical system. Leaving isn't really leaving.

We've covered the AI anxiety that's spreading through offices before, and the honest answer is that most of it is misplaced fear about job replacement. This story is different. This isn't about whether AI will take your job. This is about whether the tool you rely on to run your business has the right to hand your financial records to a third-party AI company, notify you in a ToS update, and then remove the exit ramp six months later.

The EU regulatory apparatus exists precisely for situations like this - moments where a dominant platform leverages its installed base to force adoption of new data practices before users can meaningfully respond. The Intuit-Anthropic partnership will hit European QuickBooks users differently because GDPR gives them actual teeth. US users don't have that backstop.

The "Safety" Company Problem

Here's the thing that gets me. Anthropic's whole brand is safety. Their homepage talks about building "reliable, interpretable, and steerable AI systems." They position Claude as the thoughtful, careful alternative to move-fast competitors. And maybe that's true at the model level. But at the business development level? They just signed a deal to pipe financial data from 100 million people into their ecosystem on an opt-out basis that's actively being removed.

That's not a safety company making a cautious decision. That's a growth company making a growth decision and dressing it in safety language.

I think about this the same way I think about Rey's arc in The Last Jedi - everyone keeps saying Rian Johnson broke something, but what he actually did was reveal the tension that was already there. Anthropic isn't breaking trust by doing this. They're revealing that their safety posture was always in tension with their commercial ambitions. The deal with Intuit just made the tension visible. I've explained this exact dynamic to Chris four times this week and he just nods with that calm, beautiful sincerity and says "hm, interesting" and I can't tell if he agrees or if he's just being kind. Probably the latter, honestly.

The more important question for anyone running a business is: what do you actually do now?

What Business Owners Should Actually Think About

First - and this is urgent before May 8th - if you're using QuickBooks Online, check your settings now. The AI-powered bank feed is becoming mandatory. You have a window to opt out that is closing. Use it if you want it, or at least make the decision deliberately rather than by default.

Second, think carefully about your CRM and operational data stacks. The Intuit-Anthropic deal is going to accelerate similar partnerships across every major business software category. If you use a CRM that holds your customer conversations and deal history, the question of what happens when that vendor signs an AI partnership deal is worth thinking about now, not after the fact. Same goes for payroll platforms that hold your salary data and headcount information.

Third - and I want to be direct here - the actual functionality being built is not the villain. Campfire's customers using Claude have reduced close times by three days per month on average, while cutting time spent on bank reconciliations by 90%, and reporting by 50%. Those results are real. An AI that can tell you which customers are consistently late paying, forecast your cash runway, and flag margin anomalies across fifteen restaurant locations - that's valuable software. The problem isn't the capability. The problem is the consent architecture around it.

Good technology doesn't require you to remove the opt-out. If this was genuinely better for users, users would opt in. The fact that Intuit is eliminating the choice tells you everything you need to know about how confident they are in their own value proposition.

The AI integration market is going to keep moving in this direction - more partnerships, more data pipelines, more "personalized experiences" built on your financial records. The smart move isn't to resist AI in accounting. The smart move is to demand that the companies holding your data treat consent as something more than a checked box buried in a ToS update nobody reads.

Some companies will eventually compete on exactly that - on being the accounting platform that gives you genuine control. When that product arrives, it's going to have a serious acquisition opportunity among the segment of QuickBooks users who've been clicking the X button on AI assist prompts for eighteen months and getting angrier each time.

That market is bigger than Intuit thinks it is.