Worth Founders Host a First Look at Decision Intelligence in Action

Most underwriting AI stops at the alert

It spots a mismatch, raises a flag, and hands the case back to a person who still has to manually find the owner, check the registry, and write up the decision. Meaning the queue grows, and the data stays incomplete.

In our recent webinar, "Intelligence That Acts, Not Just Advises," Worth Chief Executive Officer and Co-founder Sal Rehmetullah and Chief AI Officer and Co-founder Jacques Fu showed what happens when agents do that legwork instead. With Continuous Decision Intelligence, agents request missing information, resolve what they find, and execute decisions strictly within the rules a team has written.

Here is a recap of the live demo and the Q&A that followed, for risk and underwriting teams tired of manual review and for leaders who want lead flow to set the pace of onboarding instead. If you would rather see it for yourself, watch the on-demand replay.

Your playbook guides the decisioning

Every institution underwrites differently, so the session started with the rules. Worth takes a team's risk playbook and turns it into an automated workflow. Compliance checks, such as identity, licenses, and registrations, use one set of logic. Risk calls on fraud, credit, and portfolio exposure use another. Each playbook can shift by business type and risk tolerance.

Underneath it all, Crosswalking Technology links the scattered identifiers a business leaves behind, such as a DBA, an EIN, and an address, into a single profile. It draws on Worth’s proprietary database of 350M+ businesses across 200+ countries and territories. Pre-Fill then uses that profile to cut applicant input fields by upwards of 80%, keeping onboarding fast without loosening KYB and KYC requirements.

From there, Decision Intelligence acts. It can approve, deny, or request more information, and every outcome arrives with a clear reason code. Nothing falls outside the logic a team has defined, and each step stays traceable for audit.

Agents tackle time-consuming underwriting tasks

The live scenarios showed some of the agents working in sequence, handling parts of the review that normally land on a person's desk. For example:

• Research Agents map each submitted business onto the graph and uncover linked entities and hidden beneficial owners at a scale manual review cannot match. They also refresh data as businesses change, so a case reflects today's picture rather than last quarter's.
• Decisioning Agents apply the institution's playbook to the enriched data and return an approve, deny, or manual review outcome. Network risk and suspicious patterns, such as a trail of dissolved companies, feed directly into the decision.
• Risk Agents monitor approved businesses after onboarding and raise alerts when things shift. They automatically clear false alarms and can tune rules as the risk picture changes, sparing analysts the fatigue of chasing noise.

Automation comes in stages. Teams can start with recommendations only, move to shadow underwriting where the agent decides alongside the human without enforcing anything, and then turn on full execution once they feel confident in the results.

The Connection Graph identifies what’s behind the business

A clean application can sit on top of a messy history. The Connection Graph exists to show it. Jacques walked through a visual map that crosswalks a business identity across hundreds of databases, then lets an underwriter follow the links outward: owners, subsidiaries, headquarters, and related companies.

Patterns that take hours to piece together by hand show up on screen. A founder who has opened and dissolved several companies in a short span gets flagged. So do undisclosed risks that a manual review would likely miss. A network risk score turns those patterns into a number that helps teams decide which cases to examine first.

The same structure handles both ends of the market. A small business produces a simple graph, enriched with signals like web presence and product offerings. A large enterprise produces a layered one, with ownership hierarchies several levels deep.

Meet CODI: the conversational agent to help manage it all

CODI is the conversational layer that sits on top of the data, the agents, and the team's playbook. Risk teams can ask what happened on a case, why a decision went a certain way, and what changed across the portfolio, then get plain-language answers with links to the right tools for the next step.

It also changes how new analysts ramp. Instead of hunting through documentation, they can ask CODI to explain a term, a risk signal, or where a case stands. Teams can control how far it goes. CODI can stay in recommendation mode or take action directly, depending on company comfort level.

Built-in governance, proven at scale

A fair question came up throughout the Q&A: how do you trust an agent with real decisions? The answer was governance built in from the start.

Teams can review and test a rule's impact before it goes live, and run new logic against their historical portfolio to see how it would have performed. Shadow underwriting lets agents decide in parallel with people, which doubles as quality control on both sides. Manual overrides stay available at any point. Every case logs the evidence gathered, the rules applied, and the workflow followed, so reasoning is traceable rather than a black box.

Plus, the Worth platform connects to existing systems through API and SDKs. That flexibility helps explain why more than 50 enterprise customers, including a global payments leader, a major business credit card provider, and a leading card network, rely on Worth across more than 5 million applications a year (and counting).

See it run on your playbook

The webinar showed agents working through example scenarios, but a custom demo could show them working through yours, with your rules, your risk tolerance, and your data. Our team will walk through how research, decisioning, and risk agents could handle your applications, and where automation fits the way your team works today.

Curious to see how this could work for your company? Schedule a demo.