The short answer
“Chart.ai is the Healthcare Intelligence layer from Grelin Health. It reads a patient’s full history, reasons across it using clinical and payer rules, and suggests codes, findings, gaps and risks. Every suggestion comes with a confidence score and the exact sentence behind it. A coder, provider or reviewer accepts or rejects each one. Chart.ai never diagnoses, never writes to the chart on its own, and never replaces the person who signs.”
What Chart.ai Is
Every system in healthcare sees a slice of the patient. The EHR holds the visits it recorded. The payer system holds the claims it paid. The care management tool holds the program it runs. Each one is accurate about its own slice. None of them reads across the whole history: every condition, every result, every treatment tried, across every setting and over the years.
Coding decisions get made exactly where that gap hurts. A coder opens a note, and the note is one visit. The diagnosis that matters might sit in a specialist letter from two years ago, a hospital discharge summary, or a lab result that never reached the problem list. Nobody has the time to read all of it. So conditions get documented but never coded, and treatments get billed without the history that justifies them.
Chart.ai exists to do that reading, and to hand the result back in a form someone can check.
The five steps Chart.ai runs
Gather
Pulls in notes, letters, reports, results and outside records from the sources a customer already has. Scanned files count. So do documents from other organizations.
Assemble
Reconciles duplicates and rebuilds one timeline, so the output is a single picture of the patient across every condition, setting and year, not a stack of separate documents.
Reason
Checks that picture against the clinical and payer rules for the patient's specific problems. Those rules are Playbooks, written in plain language by certified coders and clinicians.
Produce
Returns findings, gaps, risks and suggested codes. Each one links to the exact text in the record and the exact rule that triggered it. Nothing arrives as a bare answer.
Deliver
Sends the intelligence into the tool the team already works in, where a coder, provider or reviewer accepts or rejects each suggestion.
Produce, then a human acts
Step five is the point of the whole design
100% reviewed
How a Suggestion Works
Here is the loop made concrete. The example is illustrative. It is not a real patient, but it shows the shape of the output.
A patient with type 2 diabetes comes in for a routine visit. Today’s note mentions diabetes and a medication change. It says nothing about the kidneys. Chart.ai reads further back, and in a nephrology letter from last year, one sentence states the patient has stage 3 chronic kidney disease related to the diabetes. That condition has never been coded in this practice.
What the reviewer sees
The finding
A documented condition that has not been captured or acted on.
The suggested code
Diabetes with kidney involvement, with the chronic kidney disease stage code alongside it.
The evidence
Nephrology letter, last year: the patient has stage 3 chronic kidney disease related to the diabetes.
The rule
The Playbook rule that connects the two conditions.
The confidence score
The coder reads the sentence. If it supports the code, they accept. If the letter is ambiguous, they reject it or send a query to the provider. Either way the decision takes seconds instead of a chart hunt, and it belongs to a person.
The confidence score is not a promise that the code is right. It is a signal of how strongly the record supports the suggestion. High confidence with clear evidence can be reviewed quickly. Lower confidence deserves a slower look, or a query to the provider. Because the evidence link sits next to the score, you never take the number on faith. You can read the sentence yourself.
The human stays because the work needs one. Chart.ai does the reading, which is the slow part of coding, so a reviewer moves from hunting to judging. The right answer often depends on context a system cannot see: a conversation with the provider, a policy quirk at one payer, an old note that was simply wrong.
What Chart.ai Does Not Do
Buyers get burned when a tool does something different from what they thought they were buying. So here is the plain list.
Not a claim scrubber
A scrubber works on a finished claim and judges only the result. Chart.ai works before the claim exists, on the record, where the coding is actually decided.
Not an autonomous coder
No code goes straight to the claim with nobody looking. Every suggestion is reviewed. The coder is the final step in the design, not a fallback for when software fails.
It does not diagnose
It surfaces what the record already says and what the rules require. It does not decide what is wrong with a patient. A finding is a pointer to evidence, not a clinical opinion.
It does not write to the chart
Nothing goes into the record unless a human puts it there. Chart.ai delivers suggested codes and draft wording to the tool where a coder or provider signs.
It does not replace your systems
The EHR stays. The care management tool stays. The billing tool stays. The layer connects to the history you already have and sends structured intelligence back.
Not a dashboard or a chatbot
You do not log into a screen of charts about your own data, and you do not type questions at it. The output is a finding, a gap, a risk or a code, delivered where the work happens.
One more thing it is not:
Chart.ai touches the revenue cycle, and some of Grelin’s other solutions are built for it. But Chart.ai itself is about understanding the patient. Better coding is one result of that, not the whole story.
Three Tools, Side by Side
None of these is automatically wrong. A scrubber catches real formatting errors. The point is that they do different jobs, and it helps to know which job you are buying.
Claim scrubber
When it works
After the claim is built
What it reads
The claim
What it produces
Edits and flags on a claim
Who decides
The biller, after the fact
Evidence shown
A rule name, usually
Fit with your tools
Sits at the end of billing
Autonomous coder
When it works
During coding, replacing the step
What it reads
The note in front of it
What it produces
A final code
Who decides
The software
Evidence shown
Often none
Fit with your tools
Often replaces a step
Chart.ai
GrelinWhen it works
Before the claim exists, while the record is read
What it reads
The patient's full history across sources
What it produces
Suggested codes, findings, gaps and risks with evidence
Who decides
A human reviewer, every time
Evidence shown
The exact sentence and the exact rule
Fit with your tools
Sits behind the tools already in use
Questions worth asking any vendor
Whether you pick Chart.ai or not, these questions tell you a lot about any coding tool. If the answers are vague, that tells you something too.
When in the process does it work?
Before the claim exists, or after it is already built? The answer tells you whether the tool can shape the coding decision or only inspect the result.
What does it read?
One note, or the full history across every source? A single visit cannot show the condition documented two years ago in another setting.
Can I see the evidence for each suggestion?
Not a rule name. The actual sentence from the record, shown next to the suggestion, with its date and source.
Who makes the final call?
A person or the software? And can you audit that decision later when a claim is appealed?
Who writes the rules?
And how does a change get checked before it ships? Rules should be authored by people who have worked the policies and the payers.
How is accuracy measured?
Against whose reviewers, and on whose charts? Accuracy only means something when it is measured on your own records.
What do I have to replace?
And what stays exactly as it is? A layer connects to what you run. A replacement asks you to rip something out.
The loop, in four words
Read, suggest, show the evidence, a human decides. Chart.ai produces. A person acts.
Frequently Asked Questions
What is Chart.ai in one sentence?
It is the Healthcare Intelligence layer from Grelin Health that reads a patient's full history and suggests codes, findings, gaps and risks with evidence, for a human to accept or reject.
What does Healthcare Intelligence mean?
It means software that assembles a patient's full history across every source, setting and year, reasons across it with clinical and payer rules, and hands structured output to the systems people already use. A finding, a gap, a risk or a code, each with its evidence.
Is Chart.ai a medical coding tool?
Coding is one of the things it produces. The bigger job is understanding the whole patient, and suggested codes come out of that.
Does Chart.ai replace our coders?
No. Every suggestion is accepted or rejected by a person. The goal is to remove the hunting and the rework, not the judgment.
Is it the same as a claim scrubber?
No. A scrubber checks a finished claim against generic rules. Chart.ai works earlier, on the record, before the claim exists.
Does Chart.ai diagnose patients?
Never. It points to what is already documented and what the rules require. A finding is a pointer to evidence, not a clinical opinion.
Can it change the patient record by itself?
No. It never writes to the chart on its own. A person puts anything into the record.
What systems does it work with?
It reads what the system stores, whether that is Epic, Oracle Health, athena, PointClickCare or a scanned file. The customer does not have to change their system.
How long does it take to go live?
It is measured in weeks. The layer connects to the history the customer already has, so there is nothing to rip out.
How do I know a suggestion is trustworthy?
Every one shows the confidence score, the exact text from the record and the rule behind it. You can read the sentence and check it yourself in seconds.
Related
Chart.ai: the Healthcare Intelligence layer