April 1, 2026
Every ACO leader understands the premise: spend less than your benchmark, hit quality thresholds, keep the difference. Simple in theory. In practice, the Medicare Shared Savings Program (MSSP) paid out $4.1 billion in shared savings in performance year 2024, but nearly 40% of participating ACOs still failed to generate savings at all.
The gap between ACOs that earn and ACOs that bleed isn't strategy decks or population health slogans. It's operational execution at the patient level, thousands of times per day, across fragmented provider networks. And that's exactly where AI is starting to change the math.
Let's be precise about the mechanics, because the details matter for understanding where technology creates leverage.
CMS calculates a per-beneficiary-per-year (PBPY) spending benchmark for each ACO based on historical expenditures, regional averages, and risk adjustment (HCC scores). For ENHANCED track ACOs, this benchmark is trended forward and blended with regional spending to prevent ratcheting.
The benchmark is not what you spent last year. It's a prospectively set target that CMS believes your population should cost. This is a critical distinction: your benchmark can move against you if regional costs drop or if your risk coding was aggressive in prior years.
Before you see a dollar, your actual expenditures must fall below the benchmark by a minimum savings rate. For most BASIC track ACOs this is between 2-3.9%. ENHANCED track ACOs have a fixed 2% MSR. Miss it by a fraction of a percent, and you earn nothing.
This binary threshold is where many ACOs leave money on the table. An ACO with 10,000 beneficiaries and a $12,000 PBPY benchmark needs to save roughly $2.4 million just to cross the MSR. Every dollar of avoidable spend that slips through — an unnecessary ED visit, a missed medication reconciliation, a delayed transition of care — erodes that margin.
Even if you clear the MSR, your shared savings percentage is modulated by quality performance. CMS uses the APP (ACO Performance Pathway) measure set, which includes CAHPS survey results, claims-based measures (like all-cause readmissions), and eCQMs. Poor quality scores can reduce your sharing rate from 75% down to as low as 40% in ENHANCED track.
This means quality isn't a compliance checkbox — it's a direct multiplier on your revenue. A 10-point swing in quality scores on a $5 million savings pool is a $500,000 difference in what hits your bank account.
When we analyze the spending patterns of high-performing ACOs, the savings cluster in a few predictable categories:
Inpatient spending represents 35-40% of total Medicare expenditures for most ACO populations. High-performing ACOs reduce admissions by 8-15% through better chronic disease management, timely transitions of care, and proactive identification of deteriorating patients. At an average Medicare admission cost of $13,600, preventing even 200 admissions across a 15,000-life ACO saves over $2.7 million.
SNF and home health spending is the second-largest lever. ACOs that actively manage post-acute networks — steering to high-quality, lower-cost SNFs and accelerating home health transitions — routinely save 15-25% on post-acute spend. For a typical MSSP ACO, that's $1-3 million annually.
Unnecessary ED utilization costs Medicare roughly $2,200 per visit for treat-and-release encounters. ACOs with robust care navigation programs reduce avoidable ED visits by 10-20%, but only if they can identify high-utilizer patterns and intervene before the next visit.
Out-of-network specialty referrals and duplicative imaging erode savings silently. Most ACOs don't even have visibility into referral patterns across their network until claims come back 90 days later.
The categories above are well-known. Every ACO consultant presents them on slide 12 of a strategy deck. The problem is execution at scale — which is where AI creates a fundamentally different operating model.
Legacy risk stratification uses last year's claims data and HCC codes to flag "high-risk" patients. By the time you identify them, they've already been admitted. AI-driven models that integrate real-time clinical data, ADT feeds, pharmacy data, and social determinants can identify patients trending toward an acute event 30-60 days earlier.
At PBACO, deploying AI-powered risk stratification across their attributed population identified 23% more at-risk patients than their prior rules-based model, with a 41% improvement in positive predictive value. That precision matters: care managers have finite bandwidth, and every false positive wastes hours that could prevent a real admission.
Quality scores directly affect shared savings revenue. But closing care gaps across a distributed provider network — where patients see 4-6 different providers and data lives in separate EMRs — is an operational nightmare. Most ACOs rely on manual chart chasing and quarterly gap reports that are stale on arrival.
AI-driven care gap engines process claims, clinical data, and EMR extracts in near-real-time to identify open gaps, match them to the correct attributed provider, and trigger automated outreach workflows. At Union Health, this approach closed 34% more HEDIS gaps within the first measurement year compared to their manual process, directly lifting their quality score and increasing their shared savings multiplier by an estimated 8 percentage points.
This is the overlooked lever. ACOs don't just need to reduce costs for their attributed population — they need to ensure that the care they do deliver is properly documented and reimbursed. AI-powered denial prevention that catches documentation gaps, coding mismatches, and authorization issues before claim submission recovers revenue that would otherwise leak out through denials and write-offs.
For ACOs in ENHANCED track that face downside risk, this is doubly important: every dollar of under-coded care raises your apparent per-beneficiary cost, making it harder to beat the benchmark.
Most ACOs discover referral leakage and cost outliers when claims data arrives months later. AI systems that integrate ADT feeds, referral data, and claims in real-time give ACO leadership same-week visibility into network performance. Which PCPs are referring out-of-network for cardiology? Which SNFs have the highest readmission rates? Which practices have open care gaps trending toward year-end failures?
This operational intelligence transforms ACO management from retrospective reporting to prospective intervention — and that timing difference is often the difference between earning shared savings and missing the MSR by half a percent.
CMS announced 72 new ACOs entering MSSP for the 2026 performance year. Based on what we see working across our customer base, here's what separates the ACOs that earn in year one from those that spend three years "building infrastructure":
You cannot optimize what you cannot see. The first 90 days should focus on ingesting claims feeds, establishing ADT connections with major hospitals, and building a unified patient record across your network. Skip the dashboards — focus on the data plumbing.
By month four, you need a functioning risk model running against your attributed population. Don't wait for perfection. A model that's 70% accurate and running is infinitely more valuable than a 95% accurate model that launches in month 14.
Your care managers are your most expensive and most constrained resource. Every minute they spend on manual outreach scheduling, chart review, or gap report reconciliation is a minute not spent on patient engagement. Automate the workflow orchestration so humans focus on the clinical judgment that actually prevents admissions.
Don't treat quality reporting as a year-end fire drill. Implement continuous quality measurement that tracks gap closure rates monthly, identifies underperforming practices early, and triggers interventions with enough runway to affect the performance year.
ACO shared savings is not a payer innovation exercise. It's an operational engineering problem. The ACOs that consistently earn — the ones generating $3-8 million in annual shared savings — have built systems that execute thousands of micro-interventions per week: the right patient flagged, the right gap closed, the right referral routed, the right admission prevented.
AI doesn't replace that work. It makes it possible at the scale MSSP demands. When PBACO identifies 23% more at-risk patients, or Union Health closes 34% more care gaps, those aren't AI vanity metrics. They're the operational inputs that directly determine whether an ACO clears its MSR and at what quality multiplier.
The math is straightforward. The execution is hard. And the window for the 72 new MSSP entrants to get it right is exactly one performance year.
Gautam Chowdhary is CTO of Zynix AI, where we build the data and AI infrastructure that ACOs use to operationalize shared savings. If you're entering MSSP or trying to move from BASIC to ENHANCED track, let's talk.