The stated reason determines the shape of the appeal
The stated grounds of a denial determine what an appeal must prove. An appeal against a medical-necessity denial is handled entirely differently from an appeal based on inadequate documentation. Answering the wrong kind wastes the appeal.
It is often difficult to tell which letter is which, because denial letters usually use standard phrasing that can carry multiple meanings. The exact requirement is generally referred to rather than stated.
Previous appeals are the best guidance you have, because what worked with this payer and what did not tells you more than anything stated in the policy.
The clinician always handles the clinical aspect. The agent decides what the appeal should include from an administrative standpoint and compiles the appropriate material for the clinical section. It does not prepare the clinical justification. The system is designed to prevent that.
How the agent is built in Agent Studio
Prompt skills store earlier appeal results: the reasons given, the methods used, and the outcomes. That is the institutional memory that normally sits in one person's head.
The system prompt lists the kinds of denial you may encounter and the administrative requirements associated with each. It also imposes a strict restriction: prepare the administrative sections, decide what the clinical section should cover, and never include any clinical reasoning.
The deadline and the submission method must be provided before any other information, because an appeal prepared after the deadline is of no use, even if it is of a high standard.
The Agent Chat node receives the extracted denial and the documents retrieved through the appeal process.
A Slack node sends the draft to the clinician. At that point the clinical section is still a clearly marked placeholder that indicates what it should contain.
A submission occurs only when the clinician has completed and approved the draft. Google Sheets records each appeal so the prompt skills stay current rather than showing results from two years earlier.
What this agent is built from
- Prompt skills: prior appeal outcomes by reason type, which is the memory that usually sits with one person.
- System prompt: denial categories and what each requires administratively.
- System prompt: draft administrative sections only. Never generate clinical reasoning.
- Deadline first: the window and submission route are surfaced before any drafting.
- Agent Chat (Util): receives the extracted denial and retrieved documents.
- Notion (Integration): receives the draft with the clinical section left as a marked placeholder.
- When an ActionFlow is enough: If assembling a standard appeal template with patient details is all you need, build the workflow.
Frequently asked questions
Start building AI workflows
Create a free account, open a template or a blank canvas, and run your first ActionFlow.
Newsletter
Get product updates
New nodes, agents, and product notes. We send mail only when we have something worth opening.