Compliance AI, Orchestrated: Introducing Hummingbird’s Research and Review Agents

Today, we're expanding our agent orchestration platform with a new Research Agent and a new Review Agent. Together, these two AI Agents handle the majority of routine compliance work – enabling organizations to not only increase efficiency today but also build an AI-enabled program that scales.

Compliance leaders are being pitched AI agents for everything – customer due diligence, watchlist screening, AML triage, assistance with L2 investigations. All of them promise to do the same thing: clear your queue and scale your capacity.

Implemented correctly, that’s exactly what these AI agents will do. But adopt them too quickly, and, suddenly, you’re managing a growing patchwork of AI tools – each solving a different problem, none of them talking to each other. That leaves your team responsible to glue it all together – an operational and oversight risk that most teams aren't equipped to handle.

At Hummingbird, we’ve built the industry-leading orchestration platform for compliance teams – bringing your data, workflows, and governance into a single, shared workspace. In the new AI era, this kind of orchestration becomes even more essential. Only now, it's no longer just your team that needs to be orchestrated. It's also all of the AI agents – the ones you build and the ones you buy – working alongside them.

The compliance program of the future needs a shared operating system to coordinate across agents and enable human-agent collaboration. Today, we’re taking the first step toward this future – with two new agents that work with each other and with your team:

• The Research Agent gathers and synthesizes the evidence behind every case.
• The Review Agent takes the next step, filling out the case for your team to approve.

These two agents streamline work across your compliance program and can be used to automate everything from EDD reviews to AML triage and SAR/STR filing. The agents are also designed to work wherever your team works, whether that’s in Hummingbird or an internal case management system.

One orchestration layer, two ways to deploy AI agents

We believe that AI agents shouldn't be a medley of various tools and vendors. But we also know that not every compliance team wants to work their cases inside Hummingbird, and using our AI agents shouldn’t require switching out your core systems.

That’s why we’re offering two ways to access our AI agents: in Hummingbird and over API. In both cases, the orchestration layer remains consistent. Your AI agents share the same context, comply with the same controls, and produce a single audit trail. What’s different is simply where your team sees the output:

• Within Hummingbird. Use Hummingbird as your core investigations platform. The Research Agent and Review Agent appear directly in the analyst workflow. Your AI agents and your team work in the same shared workspace.
  
  
• Over API. Keep your existing case management system and access our AI agents using our APIs. Findings, insights, case write-ups, and recommendations from the AI Agents flow back into the system your team already uses.

Now, let’s take a closer look at our new Research and Review Agents.

Get comprehensive intelligence with the Research Agent

Whether you’re conducting customer due diligence or triaging AML alerts, researching an individual or business tends to involve a lot of swivel-chairing: different systems, multiple tabs, endless data entry. But even with all this context-switching, analysts are limited by time – one person can only cover so much ground. The Research Agent changes that.

The Research Agent delivers comprehensive entity intelligence at a scale that no manual process can match. It automatically reviews and synthesizes data from more sources and faster. The result: a more complete picture of identity and risk than an analyst could have achieved on their own. The Research Agent gets to work in the background – so when an analyst opens a case for the first time, context has already been gathered for them.

Due diligence that combines your data with external intelligence

Many vendor reports come standard with limited out-of-the-box configuration. But for intelligence to be valuable, it needs to be both comprehensive and relevant. That’s why we built the Research Agent to be completely customizable. We work with your team to tailor the agent to your specific data sources, key risk indicators, and required checks.

In addition, unlike other vendors that rely solely on third-party data, the Research Agent incorporates and analyzes your internal data, too – your customers’ transaction history, behavioral patterns, and prior alerts and cases that only you have access to.

The result is a comprehensive due diligence report using sources that include:

• Web: Searches publicly accessible sources for identity information, address history, social media presence, known relationships, and adverse media.
  
  
• Internal systems: Queries your data warehouse, pulling transaction history, linked accounts, behavioral data, and anything else your team references in your internal systems.
  
  
• Prior case history: Searches Hummingbird for any prior cases involving the same subject or related entities, surfacing connections your team might otherwise miss.

Every finding includes a citation and links back to its source, so your team can verify the agent's work and maintain a detailed audit trail of precisely what data was evaluated.

Clear routine cases faster with the Review Agent

For most teams, routine cases are the norm, not the exception. Think of your low-risk or zero-tolerance typologies where the outcome is rarely in doubt. You still need to review the few that aren’t clear-cut false positives and action the ones that are zero-tolerance. But the majority of this work is time that your analysts could better spend elsewhere.

This is the work that the Review Agent is designed to handle, with the consistency and thoroughness you’d expect from a trained analyst. It analyzes the intelligence from the Research Agent and your case data, follows your institution's procedures, and fills in the review – giving analysts a complete first pass to modify or approve rather than a blank form to fill.

Because no two compliance programs are alike, neither is how the Review Agent works. It can:

• Handle reviews at any level of complexity: From enhanced due diligence to AML triage and SAR/STR filings, the Review Agent can work across your full program.
  
  
• Mirror how your analysts make decisions: Like a trained member of your team, you can configure the agent to follow your custom instructions, decision criteria, and desired workflows.
  
  
• Draw from the same data your analysts would: It can access key data, pulling from your internal systems, watchlist screening results, transaction history, and uploaded files – so it always has the context it needs.
  
  

Start with oversight, scale toward automation

Hummingbird provides a defined path to expand agent autonomy responsibly. We know compliance teams need time to evolve your operations, to develop new procedures and frameworks for agent oversight – and the ability to tailor what actions or decisions an AI agent can make based on your risk tolerance.

We’ve designed the Review Agent to follow the three stages below, allowing you to start using agents wherever you are in your AI journey:

1. Observe: Start by comparing what the AI agent would have done to decisions that your analysts are making. The Review Agent runs in parallel, so you can measure and tune agent accuracy before the agent starts producing recommendations.
   
   
2. Recommend: Surface the agent's recommendations directly in the case for your team to accept, modify, or override, with full reasoning provided for every action.
   
   
3. Automate: Once the agent's performance is proven, expand to automated case completion, with human checkpoints or QA wherever your program needs it.
   
   

Building the orchestration layer for compliance AI

As AI transforms how compliance operates, the most successful programs won't be the ones with the most AI tools. They'll be the ones whose AI works as a coordinated system – handling more work, more consistently, and to a higher standard than manual processes could. That takes more than good agents. It takes the right foundation underneath them.

That's what Hummingbird is building: the orchestration layer for compliance that brings your data, workflows, governance – and, now, AI agents into one system. Our Research Agent and Review Agent are two of our own agents built on that layer, available in both Hummingbird and over API.

Ready to see what AI can do for your program? Meet with our team to get a firsthand look at our new AI agents and agent orchestration platform. See why industry-leading banks and financial institutions, such as Raymond James, Crypto.com, DraftKings, Etsy, BILL, and more, trust Hummingbird to run their compliance programs.