Two weeks. That is how long it took for the two largest core banking and payments processors in the US to each announce a formal partnership with a frontier AI lab.
On May 4, FIS announced it is building agentic AI with Anthropic. On May 14, Fiserv announced it is building agentic AI with OpenAI. Together, FIS and Fiserv serve the vast majority of US community banks and credit unions. If your institution runs on either platform, AI is no longer an abstract conversation about the future. It is something your existing vendor is building into the infrastructure your team uses today.
This post explains what each partnership actually involves, what the differences are, and when community banks can realistically expect to see these capabilities.
What FIS Is Building with Anthropic
FIS is calling its approach "Orchestrated Intelligence." The framing is deliberate: FIS positions itself as the governed infrastructure layer sitting between your institution and AI that acts on your customers' money. Anthropic's Claude is the reasoning engine underneath. FIS owns the agents and distributes them.
The first product is the Financial Crimes AI Agent, focused on anti-money laundering investigation. In practice, this is what it does:
When an AML alert opens, the agent automatically assembles evidence across the bank's core systems before any human investigator touches it. It evaluates transaction patterns against known money-laundering typologies, surfaces the highest-risk cases for human review, and generates better-quality SAR narratives. What currently takes days can take minutes. Every action the agent takes is auditable and traceable back to source data. The human investigator makes the final call.
Two banks are in active development now: BMO and Amalgamated Bank. General availability is targeted for the second half of 2026.
Beyond financial crimes, FIS has announced a roadmap that includes credit decisioning, deposit retention, customer onboarding, and fraud prevention, all designed to run on the same platform. Revenue, according to FIS CEO Stephanie Ferris on a May analyst call, is expected to materialize in 2027.
One thing Ferris said is worth paying attention to: "Banks are looking for us to develop the agents and deploy them for them. I don't see a view where banks would own their agents at all. They're very happy to leverage our agents. It's faster for them."
That is FIS's model in plain language. You will not be building these tools. You will be consuming them through your existing FIS relationship.
What Fiserv Is Building with OpenAI
Fiserv's approach is structurally different. Rather than building a set of managed agents and distributing them, Fiserv launched an agentic AI operating system, called agentOS, on the same day as the OpenAI announcement.
agentOS runs on AWS infrastructure and gives banks three options: use agents Fiserv has already built, build your own agents, or deploy approved third-party agents from an ecosystem of ISV partners. All of it runs within a governed architecture that includes kill switches, human-in-the-loop controls, and regulatory audit record-keeping.
Six banks co-developed the early agents:
- First Interstate Bank built a commercial loan onboarding agent that reduces manual data entry and cuts cycle times
- Boulder Dam Credit Union built a report generation agent that compresses a ten-minute task to seconds
- Salem Five, City National Bank, Bank OZK, and SouthState have deployments beginning this summer
Nine ISV companies have signed on to build agents for agentOS, including Sardine for fraud, Trulioo for identity, and Sierra for customer service.
Fiserv is also using OpenAI to compress timelines on core conversions, digital migrations, and system integrations -- which are historically among the most painful and expensive projects a community bank undertakes. No specific timelines have been announced for that work, but it is a stated focus area.
agentOS is expected to be widely available in August 2026.
The Difference That Matters
Both deals follow the same underlying logic: the core processor has the data and the client relationships; the AI lab provides the reasoning capability. But the relationship each processor wants with your institution is different.
FIS is building a managed service. Your bank will use AI agents FIS built, deployed through your FIS relationship. You will not be configuring or developing anything. The upside is lower operational burden and faster deployment. The tradeoff is less control and more dependency on FIS's roadmap.
Fiserv is building a platform. Your bank can use what Fiserv builds, but you can also build your own agents or adopt third-party ones within the governed environment. The upside is more flexibility and the ability to develop institution-specific capabilities. The tradeoff is that you need the internal resources or outside help to take advantage of it.
For most community banks and credit unions, the FIS model is probably more realistic in the near term. The Fiserv model offers more upside for institutions that want to be active participants in shaping how AI works inside their operations.
What Community Banks Can Expect and When
The timeline is clearer than it has been at any point in the past few years:
August 2026: Fiserv agentOS widely available. If you are a Fiserv client, agents should be accessible through your existing relationship. The first use cases your peers have been piloting -- loan onboarding, report generation, operational analysis -- will be available to deploy.
H2 2026: FIS Financial Crimes AI Agent reaches general availability. If you run on FIS, this is when AML investigation automation becomes an option for your compliance team, not just for BMO.
2027: FIS expects these products to contribute meaningfully to revenue. This is a reasonable proxy for when the technology will have proven out across a broader set of institutions and when the next set of agents -- credit decisioning, deposit retention, fraud -- will start to come online.
What is notably absent from both roadmaps, at least in the near term, is customer-facing AI. The first wave is compliance, fraud, and operations -- which makes sense. These are high-cost, heavily manual workflows where AI can demonstrate clear, auditable ROI without requiring your institution to navigate consumer trust questions first.
The Direction the Industry Is Heading
The pattern here is not specific to FIS and Fiserv. It is playing out across the vendor stack.
Jack Henry has been building AI capabilities into its platform. Nymbus is marketing an AI-native core. The ISVs building on Fiserv's agentOS span fraud, identity, customer service, and financial wellness. The conferences your team attends in the second half of 2026 will look materially different from 2025.
What is happening, in plain terms, is that AI is being embedded into the systems community banks and credit unions already use. You will not need to independently evaluate, procure, and integrate a frontier AI model to benefit from this technology. Your vendor is doing that work. The product will arrive through the same channel as the software you already run.
That is genuinely good news for institutions with limited technology staff. It is also a reason to pay closer attention to your vendor's AI roadmap than you might have in the past, because the decisions FIS and Fiserv are making now will shape what tools you have available for the next several years.
What to Think About Now
What has your core vendor communicated about AI availability? If you have not received a direct briefing from your FIS or Fiserv relationship manager on this, ask for one. These announcements are public. The roadmap briefing should not be hard to get.
What compliance or operational workflows are currently your highest-cost, most manually intensive processes? AML investigation, SAR preparation, and loan onboarding are the first areas these tools will address. Understanding your current baseline matters when it is time to evaluate what AI actually delivers.
Do you want to build anything yourself, or do you want your vendor to handle it? If you are a Fiserv client, the agentOS model gives you the option. Most institutions will start by consuming what Fiserv builds; some will want to develop institution-specific agents. Knowing which posture you want to take before the tools arrive is better than deciding reactively.
The AI conversation in banking is no longer theoretical, and for community institutions, it is no longer a question of whether you can access these tools. Your core processor has made that decision for you. The question now is how prepared you are to use what is coming.