Webinar Recap | Credentialing: Your Clearest AI Win in 2026

Health plan technology and AI leaders have moved past the question of whether to invest in AI. Budget is committed, leadership is aligned, and the pressure now is to show results. The harder question is where to deploy first.

That was the framing Brett Dooies, Head of Product at Verifiable, opened with in a recent webinar. The session covered agentic AI, a live CredAgent demo, the ROI case for credentialing, and the build vs. buy question that most health plan AI leaders are actively navigating.

Here is what was covered.

Where agentic AI creates the most value

Brett opened by drawing a clear line between traditional automation to agentic AI, and explaining where each performs best.

Traditional automation works well for rules-based, step-by-step workflows with consistent inputs, outputs, and decision points. Most real organizational processes, however, do not work that way.
‍

Even if there's a clear starting point and end point, the middle is often rife with all kinds of nuance and decision points that traditional automation simply can't tackle, and isn't maintainable when you do try to tackle it that way.
— Brett Dooies, Head of Product, Verifiable

Agentic AI, by contrast, is designed for workflows with high variability in the middle. Brett outlined the characteristics that make a workflow a good candidate:
‍

• A clear, bounded starting point and end state
• High variability in the steps between start and finish
• A need for reasoning across incomplete or inconsistent data
• Human-in-the-loop checkpoints at key decision moments
• A requirement to produce an auditable trail of decisions
  ‍

Those things with high variability in the middle that make more traditional automation hard are actually a benefit for agentic AI.
— Brett Dooies, Head of Product, Verifiable

The key insight: having a clear start and finish lets teams build testing criteria, called evals, to measure system performance even when the path between them varies every time.

A live look at CredAgent

Rather than describing the technology in the abstract, Brett walked attendees through a live CredAgent demo.

In the traditional model, a credentialing specialist picks up a single file and works it from beginning to end. CredAgent replaces that sequential method with parallel processing. In the demo, Brett selected 10 providers with submitted applications and kicked off the credentialing process for all of them simultaneously.
‍

We're leveraging technology and compute power instead of human power. We can translate from historical precedence, which is a credentialing specialist picks up a single file and works that file from beginning to end, to running across many providers the credentialing process all at the same time.
— Brett Dooies, Head of Product, Verifiable

He showed how CredAgent works through each file, pulling provider data, checking NPI, reviewing application documents for attestations and disclosure questions, verifying primary license information, and flagging anything that requires specialist attention. In one example, CredAgent identified a yes response to a disclosure question that should have been a no, surfacing it immediately for the specialist to address rather than requiring them to scroll through the full document to find it.

Brett noted that each step in the CredAgent workflow has a defined target outcome, built and tested to achieve that outcome reliably across inputs. He also provided context on how Verifiable reached this point. Three foundational pillars: an API-first architecture for primary source verification built across all 50 states, automated workflows developed through more than 150 customer implementations, and CredAgent as the layer built on top of that foundation. He noted the expectation of at least a 10x productivity increase over traditional means.
‍

Why credentialing is ready for AI deployment right now

Janan Dave, SVP of Operations and leader of Verifiable's NCQA-certified CVO, took over to make the case for why credentialing specifically clears every bar a governance committee will set.
‍

Given our vantage point in the healthcare ecosystem, we think credentialing is a really unique opportunity for quick wins within this space for a number of reasons.
— Janan Dave, SVP Operations, Verifiable

She walked through five criteria that make credential an unusually strong AI staring point:

• Data readiness: Organizations already using integrated, automated primary source verification software are not starting from zero. The necessary primary sources are defined and accessible in digital formats.
• Governance: NCQA and other accrediting bodies provide an established compliance framework that defines what must be auditable, where humans need to be in the loop, and the confines within which innovation must occur.
• ROI measurability: Unlike some areas of clinical care, the ROI for credentialing is formulaic. Faster turnaround time, more files processed per specialist, and lower cost per packet are concrete, shareable metrics.
• Risk profile: Credentialing does not typically involve PHI, which positions the risk profile lower than clinical interventions. NCQA guidelines also require human specialist review as a final checkpoint before any file is finalized.
• Agentic AI fit: The credentialing workflow has defined inputs, a standardized output, structured decision logic, and built-in regulatory checkpoints, which is precisely where agentic automation performs best.

The ROI case: it’s about operational scale, not just speed

Janan was direct: the ROI case for AI in credentialing is not primarily about speed. It is about operational scale.
‍

Historically, I would say that improvement in credentialing has often been presented purely as gains in speed. But when we're thinking about deploying AI here, we're really talking about more of a fix for operational scale and efficiency.
— Janan Dave, SVP Operations, Verifiable

She described how Verifiable's CVO team analyzed all of the steps credentialing specialists take when credentialing a provider, mapping where time was being spent and where AI could make the most difference.

The key finding was that nearly half of credentialing applications arrive with missing information.
‍

Credentialing for a provider with an incomplete application takes our team four to five times longer than when a credentialing application comes complete. And the time spent reviewing that application, figuring out what's there and what's missing, chasing that provider for that missing information, reviewing again when they say maybe they've updated their application profile, all of that back and forth is not actually utilizing any credentialing judgment or expertise. It's data review.
— Janan Dave, SVP Operations, Verifiable

The agentic solution changes this by enabling parallel review. Rather than a specialist working through one file at a time, CredAgent can review hundreds or thousands of applications simultaneously, routing complete applications with concerns directly to specialists for immediate review, and sending incomplete applications to an outreach queue before team time is spent on them.
‍

If you have a 10-person team processing maybe 2,000 providers a month, with a solution like CredAgent, that team can now handle 10 times the volume, 20,000 providers a month, because of that parallel processing and that support from AI.
— Janan Dave, SVP Operations, Verifiable

What changes is not the hours the team puts in, but their overall capacity, while bringing labor costs down. The strategic question then becomes what the organization does with the recovered budget and how well it can absorb fluctuations in seasonal volume.

How to build the internal business case

Janan closed her section by offering stakeholder-specific framing for health plan leaders who need to socialize AI investment internally.

For CFOs and financial leaders: Credentialing turnaround time is not just an operational metric. Each day of delay is a delay in claims activity, and parallel processing shifts TAT from variable to static. Cost per packet drops significantly as one specialist can oversee close to 10 times the volume. For Medicare Advantage plans, directory inaccuracy is a Star Rating exposure, and timely credentialing and recredentialing supports directory accuracy.

For information security leaders: Credentialing data is the upstream source. When it is disconnected from directory, claims, and enrollment data, everything built downstream inherits the problem.
‍

What we've seen in practice is that the credentialing data is the upstream source, and if that data is not clean and well-managed, everything downstream is kind of inheriting that challenge. Value-based care programs, risk models, network adequacy reporting, everything that's even patient-facing, all of those things are kind of inheriting the mess from the upstream credentialing challenge.
— Janan Dave, SVP Operations, Verifiable

For COOs: As volume grows, quality and speed compete with each other. Without AI, there is no way to retain both without adding cost.
‍

The AI solution resolves this trade-off by handling that routine, structured work in parallel, which can free your specialists up for, truly, those 20-30% of files that actually present concern, need some more investigation, and need to go to committee.
— Janan Dave, SVP Operations, Verifiable

Build vs. buy: what internal teams often undercount

Brett returned to close with the build vs. buy question. He acknowledged that many health plan organizations are directing AI budget toward internal capability building, and that the instinct to build is increasingly attractive as AI tools reduce development costs.

He was clear that credentialing is not the right use case for an internal build, and walked through why.
‍

I don't know that credentialing is the best internal build opportunity in front of you.
— Brett Dooies, Head of Product, Verifiable

The factors that drive up true cost than initial estimates tend to capture include:

• Primary source integrations: There are many hundreds of primary sources and many thousands of license types across them that need to be handled and normalized. While AI accelerates this, the cost remains high.
• Compliance infrastructure: NCQA and other regulatory bodies are not static. Relying on internal teams to maintain compliance with evolving requirements adds ongoing expert overhead.
• Exception handling: Nearly half of files require some manual intervention due to incomplete or inaccurate data. The subject matter expertise required to handle this efficiently, and to reduce provider friction in the process, is significant.
• Continuous monitoring: Credentialing includes monitoring as a separate product with its own requirements and technological challenges.

The build of this type of a system may not be as daunting as it once felt, but the maintenance of this type of a system and the required experts on your team to keep a system like this running is still quite high.
— Brett Dooies, Head of Product, Verifiable

In response to an attendee question about how to handle an internal AI team that believes they can build this faster and cheaper, Brett acknowledged that unique organizational requirements are real, and that Verifiable has built configurability into CredAgent to accommodate them. But he pushed back on the framing of speed and cost as the primary question.
‍

It's not just the time to build it. It's that total cost of ownership, and understanding what's the additional IT support and staff and subject matter expertise and legal and regulatory staff needed to support an internal version of this.
— Brett Dooies, Head of Product, Verifiable

From the Q&A

Three questions from the webinar’s live Q&A are worth highlighting.

How can cost per credential be benchmarked for the first time?

Janan offered two approaches. The most straightforward is to take the monthly or annual cost of the credentialing team and divide it by the number of providers credentialed in that period. For more granularity, organizations can conduct time studies with their teams to understand how many minutes each step takes, then translate those minutes into dollars based on compensation.

What does governance sign-off for a credentialing AI pilot typically look like?

Brett said the right stakeholders depend on org structure but will generally include the CISO, COO, and CFO, as well as any chief AI officer or data and AI officer roles if they exist. He noted that if AI governance is still a maturing area, the pilot process is itself a useful opportunity to build a repeatable governance framework for evaluating any AI vendor.

Does CredAgent scan for conditionals and all require credentialing elements?

Janan confirmed that the technology reviews all required elements for credentialing across various provider types, including NPI, state licenses, board certifications, DEA, sanctions and exclusions lists, SAM, OIG, and NPDB.

‍

FAQ: AI in healthcare credentialing

‍

Why is credentialing considered a top AI use case for health plans in 2026?

Credentialing hits a rare combination of criteria: data is already structured and digitized, compliance frameworks (NCQA) are established, ROI is measured in concrete metrics, the risk profile is lower than clinical interventions, and the workflow itself is exactly where agentic AI performs best. Most AI use cases only meet some of these bars. Credentialing meets all of them.

‍

What is the ROI of AI credentialing?

The primary ROI is operational scale. A 10-person credentialing team processing roughly 2,000 providers per month can handle up to 20,000 providers per month with agentic AI support without adding headcount. Cost per credential packet drops significantly, turnaround time becomes more predictable, and specialists focus their time on the 20-30% of files that actually require credentialing judgement.

‍

Does AI credentialing meet NCQA compliance requirements?

Yes, when designed correctly. NCQA guidelines require human specialist review as a final checkpoint before any file is finalized. This is a requirement that agentic credentialing systems are built around, not built to bypass. The AI handles data review and parallel processing while specialists retain decision authority on final determinations.

‍

Should health plans build or buy an AI credentialing solution?

The case for buying is strong. Building requires integrating hundreds of primary sources across all 50 states, maintaining compliance with evolving NCQA and regulatory standards, developing subject matter expertise for exceptional handling, and sustaining monitoring infrastructure—all on top of initial development. Total cost of ownership for an internal build is typically higher than organizations estimate upfront, even as AI tools reduce the cost of the initial build itself.