Touchless Revenue Cycle: A Healthcare Revenue Cycle Automation Framework
Founder & CEO
15+ years building healthcare technology. Founder & CEO of Nirmitee.io, whose teams deliver EHR integrations, FHIR and HL7 platforms, ABDM and NHCX connectivity, and clinical AI for health systems and healthtech companies.

Updated October 2026
A touchless revenue cycle is one where routine billing work runs without staff touching it, from scheduling through collections, and people handle only the exceptions. It is a goal you reach in parts: McKinsey's January 2026 article says the touchless model will be unlocked in pieces across the front, middle and back end, and that a human will always remain in the loop.
Healthcare revenue cycle automation works best one task at a time, so this guide gives you a framework for choosing the next task, a way to measure your touchless claim rate, and an honest read of what McKinsey did and did not say. If you need the engineering done, our revenue cycle management software development team builds the integrations this work depends on.
Correction, October 2026. An earlier version of this page credited McKinsey with a "three-office framework", a $262 billion waste figure, a 2024 report and an 80 to 85 percent touchless target. We could not trace any of these to McKinsey, so they are corrected below. We also removed our function-by-function "percent automatable" scores and savings tables, because we could not source them.
What is a touchless revenue cycle?
It is a revenue cycle in which a claim or task moves to completion with no staff edit, call or portal check. We found no standard industry definition, so write down your own before you compare any numbers, including a vendor's.
People still own policy, payer rules, exceptions and audits, and their time goes to the work the rules cannot settle.
What is a touchless claim rate, and how does healthcare claims management software improve it?
The touchless claim rate is the share of claims that go from charge to final posting with no human touch. Healthcare claims management software improves it by catching errors before submission, reading payer responses automatically and sending only true exceptions to staff.
Formula: claims closed with zero staff touches, divided by all claims closed in the same period. Illustrative example, not a benchmark: if 4,200 of 10,000 closed claims had no staff touch, the rate is 42 percent.
Three choices change this number, so state them next to it:
- What counts as a touch. We count any manual edit, call or portal lookup. Count bot rework separately.
- Which claims are in scope. Split by payer, encounter type and department. A blended rate hides where the work sits.
- When you count. Count at closure. A claim that looks clean at submission can still come back denied.
The software features that move the rate are pre-bill edits tuned to your denial history, an eligibility check before the visit, automatic reading of 999, 277CA and 835 responses, statement matching and posting against the deposit, and work queues that open only when a rule fails.
What McKinsey actually says about the touchless revenue cycle
McKinsey recommends starting with the back end of the revenue cycle, then moving toward an interconnected network of agents. That advice comes from "Agentic AI and the race to a touchless revenue cycle" by Peterson, Baxi, Chan and Mollica, published 9 January 2026. The table separates what we found there from claims often attached to McKinsey.
| Claim | What the source shows |
|---|---|
| AI could cut cost to collect by 30 to 60 percent | In the McKinsey article, as McKinsey analysis. It is an estimate of potential, not a result measured at a named health system. |
| Revenue cycle costs 3 to 4 percent of revenue, over $140 billion a year | In the McKinsey article, which cites AKASA, Harris Williams, KFF and its own analysis. |
| Nearly 20 percent of claims are denied, up to 60 percent never appealed | In the McKinsey article. The 20 percent cites a KFF study of ACA Marketplace plans, so it may not hold for your payer mix. |
| $60 million to $120 million in savings | A McKinsey "if" scenario: a $6 billion system cutting cost to collect by one to two points. |
| $200 billion to $360 billion in US savings | In McKinsey's July 2023 article, covering all US healthcare spending, of which revenue cycle is only a part. |
| $262 billion wasted on revenue cycle each year | Not McKinsey. Change Healthcare, June 2017 estimated that $262 billion of 2016 hospital charges were initially denied, and that about 63 percent of denials were recoverable. That is an old figure, and it measures denied charges, not waste. |
| A McKinsey "three-office framework", 80 to 85 percent touchless target, 25 to 35 percent productivity gain | Not found in the two articles above. We removed them. We read the article text but not its exhibit image. |
A revenue cycle automation framework you can apply to any task
Ask four questions about one task: are the rules written down, is the data structured, could an error harm a patient, and can a mistake be caught and reversed? The answers tell you whether to automate, keep a person deciding, or fix something first.
This framework is our judgment from building healthcare integrations. It is not a McKinsey model or a published benchmark. It fits McKinsey's reasoning that clear rules and lower patient risk make a task a safer place to begin.
| Stage | Task | Main data route | Why people stay involved | Suggested order |
|---|---|---|---|---|
| Back end | Payment posting | X12 835, bank EFT | Unmatched payments, unusual adjustments | First |
| Back end | Denials and A/R follow-up | 835 codes, 276/277 | Clinical appeals, peer-to-peer calls | First, routine denials only |
| Back end | Underpayment checks | 835 against contract terms | Contract disputes | Second |
| Back end | Claim submission | X12 837, 277CA | Attachments, paper-only payers | Usually already electronic |
| Front end | Eligibility | X12 270/271 | Complex or multiple coverage | Before the back end if still manual |
| Front end | Prior authorization | Portals, X12 278, FHIR APIs | Clinical justification | Simple cases first, as payer APIs arrive |
| Middle | Charge capture and edits | Orders, NCCI and payer rules | Non-standard procedures | After back-end data shows the errors |
| Middle | Coding and documentation | Clinical notes | Clinical judgment, audit risk | Later, with human review |
| Front end | Scheduling and collections | Patient contact | Patient experience, financial hardship | Later, self-service first |
Why electronic is not the same as touchless
A transaction can be fully electronic and still need staff to resolve what comes back. So adoption figures show where the data is ready for automation, and they do not measure a touchless rate.
| Medical transaction | Fully electronic, 2024 CAQH Index | What it means for automation |
|---|---|---|
| Claim submission | 98% | The sending is largely done. Gains come from fewer returns. |
| Eligibility and benefit verification | 96% | Check whether your staff still re-key or re-check the answer. |
| Remittance advice | 89% | A strong base for automatic posting. |
| Claim status inquiry | 80% | Follow-up calls and portal checks remain. |
| Claim payment | 77% | Some payments still need manual matching. |
| Prior authorization | 35% | The largest gap of these six, and the one regulation targets. |
Source: 2024 CAQH Index, health plan reported volumes. The 2025 Index, as reported by AJMC, shows prior authorization at 40 percent, claim status at 81 percent and claim payment at 78 percent. We did not open the 2025 report itself.
Where should healthcare revenue cycle automation start?
For most health systems, start in the back end with denials, A/R follow-up, underpayments or cash posting. McKinsey's reasons are that this work is labor-intensive, follows clearer rules and has fewer patient-facing touchpoints, which makes it a safer place to test.
The evidence supports a different order in one case. Many guides, including an earlier version of this page, say to start with the front end because eligibility checks are simple. If your eligibility checks are still manual, we would fix them first, because a bad check creates a denial that the back end later has to work. A one-week count of denials by reason should settle which comes first for you.
From there, follow this order:
- Take one high-volume, error-prone task and baseline it for 30 days.
- Run it with people reviewing every automated action.
- Lower review only on the case types that keep meeting your error limit.
- Feed denial reasons back into pre-bill edits, so the middle of the cycle improves.
- Scale it across the organization before starting the next task. McKinsey warns against staying stuck in "pilot purgatory".
For the detail on each back-end task, see our guides to payment posting automation and AI agents for denials and appeals.
Epic remittance automation and payment posting
Whatever your EHR, including Epic, remittance automation depends on three things: receiving the 835 reliably, matching it to the bank deposit and the claim, and sending exceptions to a work queue. Confirm what your own build and clearinghouse do, because these settings vary by organization. Our X12 EDI integration team works on these feeds.
How CMS-0057-F changes prior authorization automation
CMS-0057-F requires impacted payers to implement a Prior Authorization API, generally beginning 1 January 2027. It covers Medicare Advantage, Medicaid and CHIP programs and managed care plans, and Qualified Health Plan issuers on the federal exchanges. It does not cover every payer, and drugs are excluded.
Decision timeframes began in 2026: generally 72 hours for urgent requests and seven calendar days for standard ones, with exchange plans excluded from that part. The rule puts the API duty on payers, so providers still need to connect their own workflows. See our prior authorization automation work for how that connection is built.
How to automate revenue cycle workflows without replacing your EHR
Keep the EHR as the system of record and add automation around it through standard interfaces. Use FHIR APIs where the EHR supports them, X12 for payer traffic, and vendor-approved interfaces for the rest. Every automated write needs an audit trail.
McKinsey expects most providers to need a mix of building and buying for a truly end-to-end solution. Use this guide when you compare revenue cycle automation software, tools or a healthcare revenue cycle automation platform:
| Option | Best when | Main risk |
|---|---|---|
| EHR vendor's own tools | Your workflow fits what is included | Limited reach outside that vendor's system |
| Point software | You have one clear problem, such as denials | The "patchwork" of loosely connected tools McKinsey describes |
| Outsourced revenue cycle service | You want the outcome handled for you | Less visibility into the rules and data |
| Custom build on standards | Your payer mix or workflow gives you a reason to own it | You own maintenance. Not worth it for commodity tasks |
We do not rank products here, since we have not tested them side by side and have no after-sales or ratings data of our own. When you ask which AI tools automate the full revenue cycle from coding to collections, ask each vendor for these:
- Their definition of touchless, and results on claims like yours
- An error rate after go-live, with who audited it
- How exceptions reach staff, and how every action can be reviewed and reversed
- Whether they sign a BAA, and how you export your data and rules if you leave
- Named support contacts and response times in the contract
People also search for the revenue cycle automation investments and organizational scale of named organizations, such as VNS Health, Elevance Health and One Medical, or for results from services firms such as Infosys. Our searches found no primary source with outcome data on these programs, so we make no claim about them. Ask any vendor or peer for that data directly, using the list above.
What does automating claims processing look like from submission to payment?
A claim is checked against edits, sent as an 837, acknowledged with 999 and 277CA responses, adjudicated by the payer, and paid with an 835 and a bank transfer. A process automation engine runs those steps and opens a work queue item only when one fails.
Build for repeat delivery, because files and responses can arrive twice. A safe design uses unique keys, so a repeated 835 cannot post twice. It also tests every failure path before go-live. We publish an open-source X12 clearinghouse simulator for testing these paths without touching a real payer.
How to measure revenue cycle automation, and when to expect ROI
Measure cost, time, operations and revenue together, and report each one by function. We do not publish a typical ROI timeline, because we found no reliable public benchmark, and the numbers on this page before October 2026 had no source.
| Measure | What it tells you | Watch out for |
|---|---|---|
| Touchless rate by task | How much work needs no person | Rising while errors rise too |
| Initial denial rate | Whether upstream data is improving | Payer behavior changing |
| Denial write-off rate | Revenue actually lost | Takes months to settle |
| A/R days | How fast cash comes in | Payer mix shifts |
| Cost to collect | The overall outcome | Vendor fees left out of the sum |
| Staff hours moved | Whether time went to harder work | Hours saved but not redeployed |
McKinsey names initial denial rates, denial write-off rates and A/R days as early signals, and warns not to be discouraged by a lack of immediate financial impact from a pilot. Set the baseline and the payback test before you start.
Agents across the revenue cycle: what is real today
An agent network is the long-term vision, and most current use is narrower. McKinsey describes an interconnected network of agents from patient scheduling to collections, but says most deployments so far run through third-party vendors, and that the few health systems building their own use them as discrete point solutions.
So give each agent one job, such as checking eligibility or drafting an appeal, and limit what it can change. Send low-confidence cases to a person and log every step. Our guide to agentic AI in revenue cycle management covers the design in more depth.
Staff roles shift too. McKinsey expects people to move toward training agents, reading complex outputs and handling exceptions. Be open with your team about this early.
Where Nirmitee can help
We build the integration layer that touchless work relies on: payer connections, EHR connections, work queues and audit trails. Here is what you can check for yourself:
- Open-source Da Vinci CRD, DTR and PAS servers tested with Inferno
- An open-source X12 clearinghouse simulator
- ISO 27001:2022 certification, and we sign BAAs
We are an engineering partner. We do not run billing operations, and we are not a replacement for your revenue cycle team. For a broader view, read our complete guide to revenue cycle management.
Choosing your first touchless task? Talk to our engineers about your denial data and systems, and we will tell you which task we would start with and what it would need.
Ready to scale?
Talk to our healthcare engineering team about building, integrating, and shipping faster.
Frequently Asked Questions
What is a touchless revenue cycle in healthcare?
What is a touchless claim rate?
Did McKinsey create a three-office framework for the revenue cycle?
Is the $262 billion figure from McKinsey?
Which revenue cycle functions should be automated first?
What is the timeline for seeing ROI from an AI revenue cycle implementation?
How can a health system automate revenue cycle workflows without replacing its EHR?
How do autonomous agents support revenue cycle operations?
Does automation mean eliminating revenue cycle staff?


