Vertafore's Dustin Pettit on AI Agents That Check Your Policies

Every agency has the same two chores nobody wants: keying policy data into the management system, and checking issued policies against what was quoted and bound. Skip the first and your data is incomplete. Skip the second and you are carrying E&O exposure you can't see.

Vertafore just made two Velocity AI agents generally available to go after both: the Policy Intake Agent (up to 87% faster policy processing in early testing) and the Policy Check Agent (up to 90% less checking time per line). We asked Dustin Pettit, Senior Director of Product Management at Vertafore, what the agents actually do, where humans stay in the loop, and what an agency should have in place before turning them on. (This follows our earlier interview with Vertafore's David Kallman on AI that shows its work.)

InsNerds: Policy data entry and policy checking have been agency pain points forever. Why tackle these now?

The friction these processes cause today is exactly why we prioritized them. These are two of the most fundamental processes in an agency, and the challenges go well beyond the amount of time they consume.

Despite years of progress with carrier download and automation, agencies still spend significant time manually entering policy information, particularly for complex commercial business. When that information doesn't make it into the management system, it limits what agencies can do with their data across servicing, renewals, reporting, and ultimately growing their business.

Policy checking presents a similar challenge. It's a critical quality-control process, but one that requires considerable manual effort to perform consistently.

Advances in AI have made it possible to understand complex insurance documents in different formats, transform that information into structured data, and identify where human attention is needed. We saw an opportunity to fundamentally improve both workflows while strengthening the agency management system as the system of record.

These agents also establish a foundation for what comes next. The more accurate and complete the information in the management system, the more we can automate across the policy lifecycle. The end result is freeing up human capacity so insurance professionals can spend more time on work where their expertise is most needed—building relationships, serving clients, and driving growth.

InsNerds: The Policy Intake Agent cuts processing time by up to 87%. Walk us through what it actually does when a new policy comes in, and where a human still has to touch it.

Think about a typical commercial policy. An account manager might receive a declaration page, the full policy, several schedules, and other supporting documents. Historically, they would have to read through those documents and manually enter all that information into their agency management system.

With the Policy Intake Agent, the user uploads those documents and AI takes on the heavy lifting. It identifies the relevant policy information, extracts details such as coverages, limits, deductibles, locations, vehicles, premiums, and other supported fields, and maps that information directly into the appropriate AMS structure.

It also identifies missing information and conflicting values across documents, and it does the work in a fraction of the time it takes a person today.

The user reviews the results alongside the source material, resolves exceptions, makes any necessary corrections, and approves the creation of the policy. Nothing is committed to the AMS without that human approval.

The 87% improvement comes from early testing of basic commercial policies. But what excites me is that the more complex the policy, the greater the impact. If we can say at the minimum it’s 6x faster for a single basic policy, imagine the time savings for a full package policy or a Commercial Auto policy for a fleet business with 60 vehicles and 40 drivers. That return compounds incredibly, from hours to minutes.

And the benefits go beyond efficiency. With Policy Intake Agent, our customers are getting more complete, usable data into their system without requiring someone to manually recreate information the agency already possesses.

InsNerds: When the Policy Check Agent flags a discrepancy, what happens next? Who reviews it, and how does the agency decide whether to go back to the carrier?

The important thing is understanding why the discrepancy exists and what needs to be corrected.

Let's say the AMS reflects a $2 million coverage limit, but the issued policy shows $1 million. The Policy Check Agent identifies that difference and presents the information for review.

From there, the agency professional evaluates the supporting information. Was the AMS record incorrect? Did the carrier issue something different from what was quoted or bound? Was there an intentional change that needs to be documented?

Those are decisions that require a person’s insurance expertise.

When the management system data needs correcting, the user can approve the appropriate change. When the issued policy is incorrect, the agency needs to work with the carrier to obtain a correction or endorsement. The workflow can also support follow-up tracking and automatically documenting how discrepancies were handled, by whom, when, etc.

Our goal is to dramatically reduce the time spent finding differences so the agency can spend that time resolving the issues that actually matter.

InsNerds: Policy checking is often the first thing that gets skipped when an agency gets busy. How much E&O exposure do you think is sitting in unchecked policies across the industry?

Realistically, any exposure can create risks for an agency. A missing endorsement, an incorrect coverage limit, or a deductible that doesn't match what the client expected can become a serious issue when a claim occurs.

The amount of money agencies spend outsourcing policy checking backs that up. Those that keep it as an internal task put a decent number of hours to it, and human hours only go so far.

Policy Check Agent is a major opportunity to change that risk equation. When AI can perform much of the comparison and direct attention to meaningful discrepancies, agencies have a much more practical way to check policies consistently. It also gives agencies the ability to resolve discrepancies much faster than the back and forth with a third-party integration because it keeps the process inside the system of record.

Of course, no technology eliminates E&O risk. But giving agencies the capacity to identify potential issues earlier, document their reviews, and resolve discrepancies before they become bigger problems is a meaningful step forward. When things are connected within the system of record, resolving discrepancies is faster. Period.

InsNerds: Carrier documents vary a lot in format and quality. Which lines or carriers give the agents the most trouble, and how do they handle the messy ones?

One of the realities of commercial insurance is that complexity varies significantly across policies, carriers, and lines of business.

The most challenging scenarios tend to involve policies with extensive schedules, multiple locations, specialized coverages, unusual forms, or information spread across several documents. In those cases, understanding how information relates across different sections of a policy can be just as important as extracting individual values.

There's also an important distinction between information that conflicts and information that simply isn't there. An agency may store details in their AMS that a carrier doesn't include on its policy documents. We need to account for that.

There will always be unusual documents and scenarios that require intervention. But we're building the AI agents to work across varied document structures, identify information that needs attention, and preserve the human review process. Our approach is built on a holistic canonical model that makes sure that known data gets moved and mapped to the right spots in the management system and is automatically applied to the right field in an ACORD form—which is a huge advantage.

Feedback from agencies using these AI agents helps us continually improve extraction, mapping, and discrepancy detection, particularly for those more complicated commercial workflows. And trust me—we follow-up daily!

InsNerds: How do these two agents fit with the Document Intelligence agent and the rest of Vertafore’s Velocity AI Platform? Is the idea to chain them together into one workflow?

Absolutely. That's where I believe the biggest long-term opportunity exists.

Each Velocity AI agent addresses a specific part of the insurance workflow.

Document Intelligence helps professionals find and understand information across their documents. Policy Intake extracts information from those documents and creates structured policy data in the management system. Policy Check helps validate that data against issued policies and other supporting information.

Today, those capabilities have distinct workflows and can deliver value independently.

But looking forward, as we continue developing our Agentic Agency vision, we’re innovating so those capabilities work together. Imagine a process where documents are identified, relevant information is extracted, the management system is updated with human approval, the issued policy is checked, and any discrepancies are routed to the appropriate person for resolution—all in a seamless process.

That's the direction we're working toward.

The real opportunity with agentic AI is connecting activities so that information flows through the agency without employees having to repeatedly find, interpret, and reenter the same data.

Human oversight, accuracy, and accountability will remain important throughout that process.

InsNerds: Agency staff hear "AI agent" and worry about their jobs. What did the agencies in early testing actually do with the time they got back?

Feedback from our early adopters reinforces how much time agency professionals are spending on work that doesn't use their level of expertise.

One of our early adopters works extensively with farm policies that don't come through carrier download. For that team, getting policy information into the management system has historically involved significant manual entry. They're already experiencing the benefit of having the agent handle much of that work.

Another agency described the opportunity to shift more of its team's attention toward client strategy, risk consulting, and personalized service instead of manually reviewing policy details line by line.

We're still early in measuring how agencies change their staffing and workflows over time, but those are exactly the outcomes we want to enable.

Insurance is a relationship business, and experienced agency professionals bring knowledge and judgment that technology will not replace.

I see AI as an opportunity to increase what those professionals can accomplish, allowing them to focus more of their day on serving customers, solving complex problems, and growing their book of business.

InsNerds: What should an agency principal have in place (data hygiene, workflows, staffing) before turning these AI agents on?

My biggest recommendation is to start with a clear understanding of how the work gets done today. Who is responsible for entering policies? Who reviews issued policies? What happens when a discrepancy is identified, and who follows up with the carrier?

Those are important questions because introducing AI creates an opportunity to improve the underlying workflow.

Data quality matters, but I wouldn't tell an agency it needs to complete an enormous data-cleanup project before getting started. These agents can actually help agencies improve the completeness and consistency of the information they're maintaining.

I would recommend beginning with a specific, high-value workflow, particularly commercial policies that require significant manual processing. Make sure the appropriate AMS experience is enabled, identify the employees responsible for reviewing the agent's work, and establish a clear process for handling exceptions.

Then measure the results. Look at processing time, review quality, and how effectively discrepancies are resolved.

Most importantly, involve the people performing the work. Their knowledge and feedback will be essential to getting the most value from AI.

Dustin Pettit is Senior Director of Product Management at Vertafore, where he leads innovations across multiple agency management systems to streamline operations, improve client service, and help businesses grow without adding complexity.