I've been in a lot of meetings where the settlement comes back, the contract's in a deficit, and someone asks why nobody saw it coming. I've had that meeting in consulting. I've had it on the provider side. Now I've spent the last two and a half years at Arbital Health building product to fix it.
With fee-for-service, retrospective reporting was often good enough. Risk contracts reward organizations that recognize change early and respond quickly, but most risk-bearing entities are still working with analytics built to explain what already happened rather than what's about to happen.
That blind spot grew more expensive as organizations take on more financial risk, and it's rarely obvious until the numbers come back worse than expected.
Then AI changed the math. Ask a plain-English question about your own population – which members are driving costs, which contracts are trending downward – and get a member-level answer in seconds. It feels like magic when it works. Whether it works comes down to one thing.
Here's the thing nobody shows you in a demo: AI makes bad data worse. A wrong number in a spreadsheet sits there until someone catches it. Put that same wrong number behind an AI and it comes back in a confident answer to anyone who asks. Every vendor in healthcare is selling AI right now, and every demo runs on curated data. Your data is not curated. Garbage in, garbage out. The garbage just comes out faster now.
This is where the actuarial discipline matters. When we built our Actuarial AI, the AI part is not where our time went. Most of the work went into what happens before the AI ever touches your data: validating files as they come in, checking that the data is complete and makes sense, flagging problems so you can fix them before anything goes to production. It's the boring part of the product. It's also the whole reason you can trust what comes out.
That trust gap is precisely why so much of VBC reporting and analytics is still done by hand. When nobody's confident the data feeding a system is clean, the workaround is a person checking it line by line before anyone believes the output. It's a trust problem, and it's a big part of why reporting cycles have stayed slow even as every vendor around them is selling speed.
Monthly Reporting Is Too Slow for Modern VBC
Payor and provider organizations make decisions every day that shape the financial performance of their contracts, and most of that gets tracked manually because nobody trusts an automated number they can't see inside. Patients get admitted, utilization patterns shift, risk scores evolve, coding opportunities get missed, and new high-cost members enter the population. Most organizations don't see the financial impact of those changes until weeks, or even months, later.
There's also a lag problem that people outside the actuarial world underestimate. Claims take two or three months to fully come in. So a "monthly" report is really telling you what happened a quarter ago, plus some estimates to fill in the gap. If the estimates are off, you find out even later. By the time a bad trend shows up clearly in the reports, it's been running for months.
Traditional reporting platforms were built to answer "what happened last quarter?". When you're carrying risk, the question is "what's off track right now, while there's still time to do something about it?"
By the time a financial report identifies deteriorating contract performance, the organization has already lived through whatever created it. Utilization has already occurred, and opportunities for earlier intervention have disappeared. Find a coding gap after the risk adjustment deadline and that revenue is gone for the year.
This is especially painful as margins continue to tighten and utilizations run hot across healthcare. Organizations can no longer afford to discover problems after financial performance has already slipped. They need earlier signals that allow them to change course while those changes still matter.
Visibility Should Start Before the Contract Begins
The same blind spot shows up before an organization even signs a contract.
Many providers spend months evaluating risk arrangements without having a clear understanding of their own population, projected contract economics, or benchmark performance. They commit to financial risk before fully understanding the financial opportunity or the exposure sitting on the other side of that signature
Historically, the options here were spreadsheets, one-time consulting projects, or a full platform implementation just to find out. VBC readiness shouldn't start the day a contract gets executed.
Organizations should be able to model contract economics, benchmark their populations, and validate payor data before they sign anything.
Our bar is simple: analysis that used to take a consulting project and a six-month implementation should take a day. Upload raw claims, eligibility, and revenue files in the morning. By the afternoon you can:
This is why my team and I developed Arbital Flex. You upload your claims, eligibility, and revenue data, work through the data quality checks, and then you're asking questions of your own population on the same engine our enterprise platform runs on. Our own actuarial team uses it for client work every day, and it gets better the more they use it.
That's the version that exists today. The longer arc is that payers and providers work from the same picture, and actuarial-grade answers available to any organization willing to take on risk, not just the ones that can afford an actuarial department.
Taking on risk was never the hard part in VBC. The hard part is knowing where you stand while there's still time to do something about it. AI makes the speed possible now. But clean data is what makes it trustworthy. You need both. After years of watching teams get the answer too late, I'm glad we get to build a solution to fix it.