This Week in Finance Automation

Week of February 9, 2026

Goldman Sachs is building AI agents with Anthropic to handle trade accounting and client onboarding. Alphabet is using agentic AI to process its own invoices. A new startup just raised $75M to rebuild accounting from scratch with AI. This week made one thing clear: the institutions that move the most money in the world now trust AI agents with their books.

The Big Story

Goldman Sachs Taps Anthropic's Claude to Automate Accounting and Compliance

Goldman Sachs has been quietly co-developing AI agents with Anthropic over the past six months. The target: trade accounting, transaction reconciliation, and client onboarding—back-office functions that currently require thousands of human hours.

The agents are built on Anthropic's Claude model. Goldman says it's "in the early stages" but plans to launch soon, with the goal of solving trade reconciliation and accounting matters faster while onboarding clients in less time.

Why this matters more than another "bank uses AI" headline: Goldman chose Anthropic over OpenAI. That's a deliberate choice. Claude's focus on safety, constitutional AI, and auditability makes it a better fit for regulated finance than a general-purpose chatbot. Goldman isn't using AI to generate marketing copy—they're trusting it with the accuracy of their books.

This is also the second major finance organization this week to deploy agentic AI internally. Alphabet's finance team announced it's using AI agents for invoice processing and treasury operations. When both Wall Street and Big Tech run their own accounting on AI agents, the "is this ready?" debate is over.

Source: CNBC

Our Take: Goldman picking Claude for accounting sends a clear message: accuracy and auditability beat raw capability when real money is on the line. Expect more regulated industries to follow this pattern—choosing AI models based on trustworthiness, not just benchmark scores.

Notable Developments

Accrual Launches with $75M to Build AI-Native Accounting

San Francisco-based Accrual emerged from stealth on February 5 with $75 million in funding led by General Catalyst. The platform uses AI agents as a "preparer"—organizing client inputs, identifying missing data, generating follow-up questions, and producing draft tax returns ready for professional review.

The numbers are striking: 85% reduction in preparation time and 60% reduction in review time. Every 50 complex returns processed adds capacity equivalent to one full-time accountant without any headcount increase. Early clients include H&R Block, Armanino, and Creative Planning.

Source: CPA Practice Advisor

Our Take: Accrual's approach—AI as a preparer, human as a reviewer—is the model that will win in finance. Full autonomy scares people. Augmentation that makes one person do the work of three? That sells.

88% of CFOs Now Use Agentic AI Tools

A new survey from Maximor reveals that 88% of CFOs use at least one agentic AI tool for finance and accounting tasks. But the adoption isn't evenly distributed: 27% have automated 50-75% of their workload, while 51% are still in the 25-50% range. Only 1% have automated more than 75%.

The trust gap remains real. Just 14% completely trust AI for accurate accounting data, and 86% have experienced hallucinated or inaccurate data at least once. Two-thirds say human oversight is "extremely or very critical."

Source: Accounting Today

AI Reshaping Fraud Prevention in AP

AI-powered fraud detection in accounts payable is moving from periodic audits to continuous monitoring. New systems use behavioral analysis and pattern recognition to flag anomalies in real-time—catching duplicate invoices, unusual vendor patterns, and suspicious transactions before they hit the ledger.

The shift matters because traditional fraud detection happens after the fact. By the time a quarterly audit catches a duplicate payment, the money is gone. Real-time AI monitoring catches it before the payment is approved.

Source: oAppsNet

Quick Hits

Alphabet Finance AI: Alphabet's finance department is using agentic AI to automate invoice payment and reconciliation, plus treasury functions. When Google runs its AP on AI agents, that's validation. (CFO Dive)
pleasefix.ai Raises $2.1M: The Brussels-based startup automates finance tasks inside Excel and PowerPoint using plain English instructions. Claims 10x faster task completion. Pre-seed led by Pitchdrive. (Business 2.0)
AI Close Time Impact: 31% of CFOs believe AI will reduce monthly close time by at least five business days. That's a week. For finance teams stuck in 10-day close cycles, AI could cut that in half. (Accounting Today)
Basware + Redmap: Basware's acquisition of Australian AP automation provider Redmap gives it mid-market reach in APAC, where the AP automation market is projected to double from $770M to $1.4B by 2030. (Basware)

Numbers of the Week

Metric	Value	Context
88%	CFOs using agentic AI	Only 12% haven't adopted any AI tools
$75M	Accrual's launch funding	Largest AI-native accounting raise this year
85%	Prep time reduction (Accrual)	AI agent as preparer, human as reviewer
14%	CFOs who fully trust AI accuracy	86% have experienced hallucinated data

What We're Watching

Goldman's Claude deployment timeline. Goldman said "soon" on launching its accounting AI agents. When the world's most prominent investment bank goes live with autonomous trade reconciliation, it will set the standard for what "production-grade finance AI" looks like. Every AP vendor will benchmark against it.

The trust gap at 14%. Only 14% of CFOs completely trust AI for accounting accuracy, yet 88% are using it. That gap is either going to close fast as the technology improves—or it's going to create a new category of "AI audit" tools that verify AI-generated financial data. Either way, it's a business opportunity.

Accrual's impact on AP automation. Accrual launched for tax prep, but their AI-as-preparer model applies directly to invoice processing. If every 50 invoices equals one headcount saved, the ROI math gets compelling fast for mid-market AP teams.

The Bottom Line

This was the week Wall Street said "yes" to AI agents in accounting. Goldman Sachs trusting Claude with trade reconciliation, Alphabet running invoices through AI agents, and 88% of CFOs using agentic tools—it all points to one conclusion: the trust barrier is falling.

But falling isn't the same as gone. With 86% of finance leaders having experienced AI hallucinations and only 14% fully trusting the output, the winning strategy is still human-in-the-loop. Build AI that does the prep work. Let humans do the judgment work. That's the model Accrual is betting $75M on, and it's the model that makes sense for AP automation too.

For finance teams still processing invoices manually: your peers are automating 25-75% of their workflow. The question isn't whether to start—it's how far behind you're comfortable being.

Want This Week in Finance Automation in your inbox every Sunday? Join the waitlist.

Previous Editions

This Week in Finance Automation: Intuit Bets $100M on OpenAI (Week of Feb 2, 2026)

This Week in Finance Automation: Goldman Sachs Picks Claude for Accounting AI | Week of Feb 9, 2026