AI is Already in Finance. Most Teams Aren't Ready or Capable.

Finance leaders are under pressure from every direction. Boards want faster closes and leaner operations. PE sponsors are comparing portfolio company efficiency metrics. CFOs are being asked to scale revenue without scaling headcount.

Everyone knows the AI conversation is happening; most finance teams still don’t see how far the technology has already moved.

While finance leaders are still “exploring” AI, a growing number of portfolio companies are already running core workflows on it. Invoice categorization, reconciliation research, accrual identification, document organization for month-end close. These are live operations that can run on Claude Opus 4 and Cowork.

The same technology Goldman Sachs hired Anthropic to deploy for accounting and compliance automation is available to any company that figures out how to use it.

Until recent weeks, QuantFi operated almost entirely on a team level. We ran finance departments as a service using experienced professionals across every function.

We watched AI capabilities evolve. We tested tools. We stayed skeptical.

Then something shifted. The technology crossed a threshold where the question changed from “if” to “now.”

We’re introducing AI agents into our workflows across client engagements. Not as a replacement for our team, but as a force multiplier that eliminates the manual research and document hunting that bogs down month-end close.

If we’re crossing that line, it’s a signal that ‘wait and see’ is no longer a safe posture.

We made this decision because we’re practitioners running live finance operations. We see exactly where time goes. And we hit the point where not deploying AI would mean delivering slower, more expensive service than what the technology now enables.

Claude Opus 4 and Cowork handle what we call closed-loop workflows. Tasks where the environment is defined (inbox, Google Drive, specific folders), the context is contained, and the output is reviewable.

Document-heavy research: Find the vendor contract that explains this charge. Pull all invoices for Project X from emails and Drive. Identify which expenses should be accrued based on this email thread.

Exception investigation: Why does this bank rec have a $4,200 difference? Which transactions from Ramp aren’t in NetSuite? What’s the backup for this unusual GL entry?

Categorization and organization: Split this invoice across three departments and tag by project. Categorize 50 transactions based on vendor and description. Organize files by month, entity, and document type for close.

These tasks consume 40-60% of time in month-end close, AP, and reconciliation workflows. They’re the reason close takes 10 days instead of 5.

But here’s the question that follows: if these workflows compress so much, and we automate 4 other workflows with similar compression, do we still need the same headcount?

That’s the question finance leaders are asking right now. It’s the question we’re asking ourselves as we deploy AI agents into our own operations.

1. Closed-Loop Workflows (Claude handles now)Document organization and extraction. Research across files, emails, and folders. Transaction categorization. Exception identification. This is what’s compressing right now at Goldman, at QuantFi, at leading portfolio companies.

2. System-Dependent Workflows (On the Horizon)Journal entry creation. Payment processing. Multi-step reconciliation execution. AI can do this work today, it just can’t access all the systems yet. As authentication barriers fall, this work becomes increasingly automatable.

3. Judgment-Dependent Workflows (Stays human)Strategic decisions requiring company context. One-off situations with no precedent. Stakeholder communication. Policy interpretation in ambiguous cases.

If your team is primarily anchored in Category 1 work, they’re in the compression zone right now. If they’re split between 1 and 2, they have months to shift before Category 2 compression accelerates.

Goldman didn’t get special access. They hired Anthropic to deploy the same Claude Opus 4 and Cowork that’s available to every company. We’re using the same technology at QuantFi.

Most companies approach this wrong. They pilot an AI assistant, it runs in parallel to existing workflows (no time savings), produces output no one knows how to use, and after 90 days someone shelves it.

The companies getting this right mapped workflows first. What actually happens during month-end close? Which tasks are closed-loop vs. system-dependent vs. judgment-dependent? Where does context live? Which team members spend 60%+ of their time on closed-loop work?

You have access to the same technology Goldman is using. The same technology we’re deploying across our client operations. The question is whether you know your workflows well enough to deploy it.

By the time something is proven in finance, the early movers have already rebuilt their workflows, retrained their teams, and reset productivity expectations.

Goldman isn’t waiting for more proof. QuantFi just crossed the line from evaluating to deploying. The companies that wait for certainty will spend 2027 explaining why they’re behind.

Some portfolio companies are closing in 5-6 days with 2 FTEs. Most are still closing in 10+ days with 4 FTEs, manually researching exceptions, adding headcount as revenue scales.

That gap compounds. Faster close cycles mean tighter board oversight. Lower finance costs mean better unit economics at exit.

The finance professionals in the lagging group will be defending their roles when someone asks why they’re still doing this manually.

At some point in the next 12 months, finance leaders will have a hard conversation with their teams.

It will sound like: “We’re redesigning workflows to use AI for document research, categorization, and exception investigation. Here’s what we need you to focus on instead.”

Or: “Our peer companies are closing faster with smaller teams. We need to figure out how to do the same.”

The teams that have already mapped workflows, deployed AI, and retrained their people will have that conversation from a position of strength. The teams that waited will have it from a position of defense.

We’re having that conversation with our own team right now. We’re not waiting for perfect certainty. We crossed the threshold where the technology works reliably enough to deploy, and we’re acting on it.

The question: Will your team be ready when that conversation happens, or will they be scrambling to prove they’re still necessary after the market has moved on?

If you want a structured assessment of which workflows are compressible and how to redeploy capacity before the next budget conversation, we can help. We’re not consultants advising from the sidelines. We’re operators deploying this technology in live finance operations right now.