On May 4, FIS and Anthropic announced an AI agent that performs anti-money-laundering investigations for banks. Not a dashboard. Not a copilot that suggests next steps. An agent that assembles evidence across core banking systems, evaluates activity against known typologies, surfaces the highest-risk cases, and compresses what used to take hours into minutes. BMO and Amalgamated Bank are among the first to deploy it. Broader availability is planned for the second half of this year.
One day later, on May 5, OpenAI and PwC announced they are building what they call the first AI-native finance function at enterprise scale. The collaboration targets the core operating rhythms of finance: forecasting, planning, reporting, procurement, payments, and treasury. The role of finance professionals in this model shifts from executing processes to overseeing and improving agents.
Two announcements. Two of the most cautious, compliance-heavy, process-bound functions in business. Same week.
If you have been telling your team that AI agents are not ready for your industry, this is the week that argument stopped working.
What Actually Happened
The FIS partnership is not a proof of concept. Anthropic embedded forward-deployed engineers with FIS to co-design the agent and transfer knowledge so FIS can build and scale additional agents independently. The roadmap already extends beyond financial crimes into credit decisioning, deposit retention, customer onboarding, and fraud prevention. This is infrastructure, not an experiment.
The OpenAI-PwC collaboration is similarly operational. They are building a procurement agent inside OpenAI’s own finance organization first, then applying what they learn to additional agents across core finance workflows. PwC is not selling a product. They are redesigning how finance departments operate when agents handle the process work and humans handle judgment, controls, and outcomes.
Both partnerships share a pattern worth noticing: neither one treats agents as a feature bolted onto existing workflows. They treat agents as the new foundation of the workflow itself.
Why Banking and Finance Matter
Every industry has its reasons for waiting. Healthcare has HIPAA. Legal has privilege. Insurance has actuarial standards. Government has procurement rules.
Banking and finance sit at the top of that list. AML compliance alone carries penalties in the hundreds of millions. TD Bank paid $3.09 billion in 2024 for BSA violations. The regulatory surface area covers every transaction, every customer, every cross-border transfer. If any industry had earned the right to say “we need more time,” it was this one.
They did not say that. They deployed agents.
The FIS-Anthropic agent does not bypass compliance. It performs it. It operates inside the regulatory framework, under human oversight, generating Suspicious Activity Reports that meet the same standards human investigators produce. The difference is speed and consistency. Banks currently deal with massive false-positive rates in their transaction monitoring. The agent evaluates every alert the same way, every time, without fatigue, without drift.
That is what agents look like in practice. Not replacing judgment. Replacing the manual steps around judgment so that human attention goes where it actually matters.
The Excuse Was Always About Comfort, Not Risk
When leaders in other industries say “our environment is too complex for AI agents,” what they typically mean is: we have not figured out how we would manage them. That is a real problem. It is an operations problem, not a technology problem.
FIS and Anthropic solved it by embedding engineers who co-designed the agent with the people who understand the domain. OpenAI and PwC solved it by building inside their own organization first, learning what works before scaling it. Both chose operational discipline over waiting for perfection.
The organizations that are still waiting are not being cautious. They are losing ground. Every month that passes, the teams deploying agents build operational muscle that cannot be acquired later through a rushed implementation. They learn what “agent oversight” actually means. They develop the management habits. They discover where agents add the most value and where humans remain essential.
That knowledge compounds. And the teams without it will be trying to build it from scratch while competing against teams that have been refining it for years.
What This Means for Your Team
The two questions worth asking this week:
First, what happens if you wait another six months? The answer changed. Six months ago, regulated industries could point to the absence of real deployments. That absence no longer exists. BMO has an agent running AML investigations. PwC is redesigning finance operations around agents. Six months from now, the list will be longer. The standard will be higher. The gap between organizations that started and organizations that waited will be wider.
Second, how long before this becomes the expectation and not the exception? Based on the velocity of these partnerships, not long. FIS is already scoping agents beyond financial crimes. PwC is already extending beyond procurement. The roadmaps tell you where this is going. The only question is whether your team will be building that operational capability now or scrambling to catch up later.
The hardest rooms in business just opened their doors to agents. The rooms your team operates in are not harder than banking compliance. They are not more regulated than AML. They are not more process-dependent than enterprise finance.
The last excuse is gone. What remains is the decision.