Healthcare value chain, 2026-2028
AI hits healthcare's 17 activities on different timelines. Benefits verification is automatable today. Clinical decision support needs 5 more years of trust-building. Between those two endpoints sit 15 activities, each with its own readiness curve, its own causal dependencies, and its own margin impact. The organizations that sequence correctly compound returns. The ones that try everything at once compound waste.
Diffusion takes time. The plan below maps 24 months of causal sequencing across 17 healthcare activities.
US healthcare runs on $4.5 trillion. Every dollar moves through 17 distinct activities from patient registration to final collection. The profit distribution across those activities has been stable for two decades: high-margin administrative functions (claims adjudication at 18%, denial management at 20%, charge capture at 15%) cross-subsidize thin-margin clinical operations (care delivery at 6%).
Eight of these 17 activities are pattern-matching functions: read input, apply rules, produce output. Benefits verification, prior authorization, coding, claims submission, claims adjudication, denial management, payment posting, collections. Their margins are high because the specialized knowledge required to do them was expensive to hire.
That specialized knowledge is now replicable by AI. The activities that justified their margins through complexity are the ones most exposed to displacement.
Defining condition
Margin concentration in pattern-matching activities is at its historical peak. The incumbents have never been more profitable or more exposed.
$4.5T
Annual US healthcare spend
17
Value chain activities traced
8
Activities facing AI displacement
18%
Claims adjudication margin (highest)
Profit pool map, current state
Bar width = revenue share. Bar height = operating margin. The tallest bars on thin slices are where AI hits first.
Documentation feeds coding. Coding feeds claims. Claims feed collections. Quality scores feed reimbursement rates. Each upstream improvement compounds through every downstream activity. Run it in the right order and each phase gives you the data quality, staff capacity, and institutional confidence for the next.
Run it out of order, automate claims submission before fixing documentation, and you scale bad inputs at machine speed. Denial rates climb because the AI submits claims built on the same vague notes humans struggled with.
Start where AI is already production-grade and clinical risk is lowest. Benefits verification: move to 80%+ zero-touch eligibility checks via payer API integration. AI reads complex benefit structures faster than staff. The manual process exists because plan documents are Byzantine, not because the task requires judgment.
Prior authorization: automate submission for routine procedures using AI that matches clinical notes to InterQual/MCG criteria. CMS and commercial payers are moving to electronic prior auth. Payment posting: auto-post standard remittances from ERA files. Already 70-80% automated at most systems. Push to 95%.
These three activities free the staff capacity you need for Phase 2. They also build the institutional muscle for AI adoption: governance, monitoring, exception-handling workflows. In parallel, pilot ambient documentation with 5-10 providers in one specialty to establish a quality baseline.
Phase 1 is about freeing capacity and building confidence. The margin impact is small. The unlocked capacity is the asset.
What to measure
Benefits verification zero-touch rate above 80%. Prior auth cycle time cut 60%. Payment posting auto-rate at 95%. Staff capacity freed for redeployment to Phase 2.
80%+
Target zero-touch verification rate
-60%
Prior auth cycle time reduction
95%
Payment posting auto-rate
2-4 mo
Time to ROI on Phase 1
Profit pool after Phase 1
Benefits verification and prior auth margins drop. Payment posting compresses to 4%. The rest of the pool is unchanged. Phase 1 is surgical.
Benefits verification: payer API platforms (Availity, Waystar, Experian Health) with AI interpretation layers. The API exists; the intelligence layer that reads plan complexity is the differentiator. Prior authorization: AI-powered ePA platforms (Cohere Health, Olive AI, Infinitus). CMS is pushing electronic prior auth. Align with the regulatory direction.
Payment posting: most EHR-integrated RCM suites handle ERA auto-posting. The gap is in exception detection: AI that flags underpayments against contracted rates before they post. This is where the revenue recovery sits.
Ambient clinical documentation is the highest-impact, highest-satisfaction AI use case in healthcare. Roll it across primary care and high-volume specialties. Abridge integrates deepest with Epic. Nuance DAX fits Dragon Medical environments. Heidi Health works without EHR integration for a fast start.
Connect documentation output directly to AI-assisted coding. Structured notes feed code suggestion, with human auditors on every chart during this phase. Retrain scribes as QA auditors. Reduce CDI program scope as AI pre-structures notes.
This is where compound returns start. Every documentation improvement lifts coding accuracy, which lifts clean claim rates, which lifts collection speed. First-pass coding accuracy should climb from 70-80% to 90-95%. CDI query rates drop from 15-25% to 5-8%.
Phase 2 is where the cascade begins. One improvement at the documentation layer compounds through five downstream activities.
What to measure
First-pass coding accuracy above 90%. CDI query rate below 8%. Physician documentation time cut 70%. Clean claim rate improvement visible within 60 days of coding accuracy lift.
90-95%
Target first-pass coding accuracy
5-8%
Target CDI query rate (from 15-25%)
-70%
Physician documentation time reduction
3-8%
Revenue uplift from coding cascade
Profit pool after Phase 2
Clinical documentation margin rises to 6% (revenue multiplier effect). Charge capture drops from 15% to 10% as AI coding reduces specialist labor. Care delivery ticks up to 7%.
Ambient documentation is technically ready. The bottleneck is physician trust. Doctors have been burned by EHR promises before. Dragon Medical was supposed to eliminate typing, and it created new workflow overhead instead. The physicians who adopted early are enthusiastic (Abridge reports 90%+ satisfaction). But the majority-adoption curve requires institutional evidence, not vendor demos.
That is why Phase 1 matters: it builds institutional AI confidence on low-risk activities before asking physicians to trust AI with their clinical notes. The governance frameworks, monitoring dashboards, and exception-handling workflows from Phase 1 transfer directly to Phase 2.
The US employs roughly 25,000 medical scribes. Their job is real-time transcription, sitting in the exam room, typing what the physician says and does. Ambient AI does this faster and cheaper. But the scribe does not disappear. The role transforms from transcription to QA: reviewing AI-generated notes for accuracy, completeness, and clinical nuance the model may miss.
The QA role requires higher judgment, pays better, and scales differently. One auditor can review notes from 8-10 providers instead of scribing for one. Fewer people, higher per-person value, different skill set.
With cleaner documentation and accurate coding upstream, the claims layer is ready. First-pass clean claim rates should now exceed 95% because the root causes of denials, documentation gaps and coding errors, are fixed at the source.
Claims adjudication margins compress from 18% to 12%. Not because payers are losing money, but because AI auto-adjudicates routine claims and the human reviewers left are handling genuinely complex cases. Denial management shifts from volume appeals to complex exception handling where recovery per claim is $2K-50K.
Collections sees the sharpest efficiency gain. AI prioritizes by propensity to collect, modeling patient financial behavior, insurance secondary coverage, and payment plan likelihood. AR days compress 5-10 days. The revenue impact from Phase 2's coding cascade is now visible in the financials: 3-8% uplift.
Phase 3 is where the margin restructuring shows up in quarterly earnings. The upstream improvements from documentation and coding flow through to every downstream activity.
What to measure
Clean claim rate above 95%. Denial rate below 5% (from 10-15%). AR days compressed 5-10 days. Revenue impact from documentation cascade visible in financials.
95%+
First-pass clean claim rate
<5%
Denial rate (from 10-15%)
-5-10
AR days compression
12%
Claims adjudication margin (from 18%)
Profit pool after Phase 3
Claims adjudication drops to 12%. Denial management to 14%. Care delivery rises to 8%. The margin inversion is underway.
$262 billion in claims are initially denied each year. The denial rate runs 10-15% across the industry. The root causes are consistent: missing documentation (30%), coding errors (25%), eligibility issues (20%), authorization gaps (15%), and other (10%). Phases 1 and 2 fix the first four categories at the source.
When denial rates drop below 5%, the denial management function shrinks but does not disappear. The remaining denials are genuinely complex: experimental treatments, out-of-network disputes, coordination of benefits issues. These require human judgment and negotiation. The economics flip: fewer denials to work, but each one is worth focused attention.
Care coordination and clinical decision support come last because they need the deepest integration and the most institutional trust. Connect care coordination to AI-driven risk stratification using the structured data now flowing from documentation and coding. Identify high-risk patients early. Panel sizes increase 2-3x with AI-assisted management.
Clinical decision support: move from rule-based alerts (90% override rate) to contextual, patient-specific recommendations. This only works after documentation quality is high enough for the AI to reason about the patient accurately. Quality and compliance: wire MIPS/MACRA reporting into real-time gap detection. Quality scores improve. 2-9% of Medicare revenue swings on these scores.
By month 18, you have 12+ months of data proving the full cascade. The playbook is an asset: the vendor configuration, the retraining programs, the cascade monitoring, the platform layer. It works for any health system with a similar EHR footprint and payer mix. There are 6,000+ hospitals and 200,000+ practices in the US. Most share three EHR platforms. A playbook proven at one mid-size Epic shop works at a hundred others.
The system you built is a deployable asset. Take it to peer organizations. Co-own the deployment. Compound returns across the industry instead of hoarding a single-site efficiency gain.
What to measure
12+ months of cascade data. Quality scores up 2-9%. Care coordination panel sizes 2-3x. The proven playbook ready for peer deployment.
2-9%
Medicare revenue swing from quality scores
2-3x
Care coordination panel size increase
12+ mo
Cascade data proving the playbook
6,000+
Hospitals the playbook can deploy to
Profit pool after Phase 4
Care delivery at 9%, clinical documentation at 8%, care coordination at 7%. Claims adjudication compressed to 10%. The margin inversion is complete.
The playbook you built is not a slide deck. It is a running system with 12+ months of attribution data: documentation time savings, coding accuracy improvement, clean claim rate gains, denial rate drops, quality score improvement, revenue uplift. Every metric attributed to a phase, a tool, a workflow change.
We deploy it to peer organizations as co-owners, sharing the margin uplift, licensing the monitoring platform, compounding returns across every deployment instead of billing hours at each one. The system compounds. The consulting engagement does not.
Your EHR footprint, payer mix, current denial rates, documentation burden, coding accuracy, and quality scores. A principal reads these before walking in. You walk out with the phase plan: which activities to rebuild first, which tools to evaluate, and where the margin compounds.
We make the full system work for one organization. Then we take the proven playbook to peers as a partner, co-owning the deployment, sharing the margin uplift, compounding returns across the industry.
Our upside and yours compound on the same axis. That is the only alignment that holds.