Introducing Candid AgentKits: Agentic RCM solutions that eliminate work
Today, we’re launching Candid AgentKits — a suite of agentic solutions that combine deterministic rules, agents, and insights to autonomously resolve and prevent issues in your existing billing system.
Each AgentKit targets a specific RCM workflow. The Denials AgentKit, which investigates, resolves, and prevents denials, is available now. The Eligibility AgentKit, which confirms coverage, handles exceptions, and prevents rejections and denials, is coming in early 2027. Future AgentKits could take on workflows like stale claims, contract management, and payment posting.
Why we built AgentKits
Candid Health has spent the last seven years building a touchless RCM platform that eliminates manual billing work. When designing our platform, we followed two principles. First, submit claims right the first time, instead of cleaning up issues after the fact. Second, when issues do arise, find the root cause and prevent them from recurring.
The Candid platform is now trusted by over 200 healthcare organizations and delivers a touchless claim rate of over 95% — a metric that measures the percentage of claims submitted and adjudicated correctly the first time with no manual work.
Our existing customers rely on the Candid platform as their system of record for end-to-end RCM workflows, but we know not every team is in a position to change systems. Those teams shouldn’t have to wait until a contract ends or retrain their teams on a new tool to start eliminating manual work.
That’s why we created AgentKits. Designed to plug into your existing billing system, they use data engineering, automation, and AI to solve RCM’s toughest problems. When building AgentKits, we drew upon all the hard-won lessons we’ve learned from building the Candid platform.
Using payer phone calls, eligibility checks, and policy analysis, the Denials AgentKit pinpoints root cause of each denial. Then it prepares appeals, handles write-offs, and routes exceptions to billers. The Eligibility AgentKit verifies coverage before care, generates patient estimates, and handles exceptions with payer calls, policy analysis, and plan carve-out identification. Both AgentKits identify trends and patterns, automatically suggesting upstream rules that fire as far upstream as at scheduling to prevent errors before they occur.
How agent-only solutions fall short
There’s been a proliferation of agentic solutions in healthcare billing. But those that rely solely on agents, however impressive in demos, fall short in production.
First, agent-only solutions throw agents at every task, even ones that don’t require you to burn tokens. Once you understand the root cause of a class of issues, you don’t need agents to reason from scratch to resolve them. But agent-only solutions have no way of codifying learnings from investigating and resolving issues. Instead, they waste tokens on expensive agent runs to reinvestigate issues you already know how to fix.
Second, agent-only solutions work with limited context. Because point solutions don’t come with a broader data platform, they’re missing valuable information like historical claim, patient and remit data, as well as a comprehensive view into the organization including providers and credentialing spans. Partial data inevitably leads to denials.
Finally, agent-only solutions are focused on automating manual work, not preventing work altogether. At best, they clean up denials, but even a successful appeal carries rework costs, delayed cash, and patient friction. And because they bill per task, every denial is more revenue. These solutions simply aren’t incentivized to stop issues from recurring.
AgentKits deliver reliable results at lower cost
Over the years, Candid Health has worked to deeply understand RCM and biller needs, and to co-build the product with customers. Our experience has taught us that an agent-only approach is not effective.
In AgentKits, agents are just one component of a full-stack solution. AgentKits combine comprehensive data on each claim, deterministic rules that automate known workflows, and LLMs and ML techniques that detect recurring issues and turn them into preventative upstream rules. Compared to agent-only solutions, AgentKits get more claims right the first time, at a lower cost.
Connections to your data
Each AgentKit integrates bidirectionally with the tools powering your RCM workflow, such as your EHR, clearinghouse, and patient scheduling system. This creates a unified data foundation that gives agents all relevant context behind each issue.
Rules and agents work together
AgentKits start with a rules layer that codifies your processes, your know-how, and payer logic reverse-engineered by Candid. An orchestration layer then routes each issue to the most reliable, cost-effective tool for the problem.
For issues that meet known criteria, rules act automatically and instantly, with no token spend. For more complex work, agents take on triage and complete tasks like calling payers, checking portals, and reviewing payer policies. Because healthcare data is messy, agents cross-check multiple sources before drawing conclusions, and then take action directly in your billing system.
Lastly, true exceptions go to your billers in the workflow tool they already use, so they can approve an agent's proposed fix or take action themselves.
Continuous improvement
AgentKits use LLM and ML techniques to analyze claim history, adjudication outcomes, and the actions your billers and agents take. When they detect recurring patterns and shifts in payer behavior, they turn those insights into new upstream rules that prevent issues from occurring again. The more you use AgentKits, the smaller your agent workload and associated costs become. Learnings are also shared across AgentKits to further improve overall RCM performance. For example, when the Denials AgentKit traces a pattern back to eligibility, that insight is turned into a pre-service rule. That rule can be picked up by the Eligibility AgentKit to inject coverage-related warnings for that same plan into the scheduling workflow for the next patient.
Radical visibility
You can’t deploy agents autonomously and with confidence if you can’t observe or audit what they’re doing. AgentKits keep granular audit logs and activity history. Users can see what decisions were made and why, and whether a rule or agent took the action.
The Denials AgentKit in action
To show how the Denials AgentKit works, let’s walk through how it handles a notoriously messy problem: plan carve-outs.
A claim is denied even though the patient’s eligibility check showed that they were covered. The denial arrives with a generic “non-covered benefit” reason that says nothing about why the service wasn’t covered. Since no single data point gives you the full picture, the AgentKit orchestrates a multi-source investigation. It runs an eligibility check to parse plan details and reviews payer policies to extract plan-level exclusions. Because these sources are often outdated or incomplete, it also pulls data on past adjudications directly from the practice’s billing system to see how the payer has actually behaved historically. If it still can’t draw a definitive conclusion, an agent calls the payer and asks directly.
Here, the AgentKit identifies the cause of the denial as a carve-out: the service is covered by a separate payer, not the one originally billed. It prepares a new claim based on the carve-out payer’s billing rules, such as adding modifiers, and submits it.
The AgentKit then turns what it’s learned into a new pre-submission rule. The next time a claim comes through for that employer group and service, the rule fires instantly, updating the payer directly in your billing system and routing the claim to the right payer before submission. Because a rule now handles the routing, this specific type of denial is permanently prevented, with no additional token spend.
AgentKits: A new path to touchless RCM
With AgentKits, we’re bringing Candid’s touchless RCM to more organizations, with no rip-and-replace of billing systems required. Like the Candid platform, AgentKits are built to eliminate issues, not profit from them. The more you use them, the fewer issues you have, so agent workloads and your costs decrease over time.
The Denials AgentKit is available now, the Eligibility AgentKit will be available early 2027, and more solutions will follow soon. We’re excited to keep building toward a future where getting paid in healthcare is effortless.