The math,
run honestly.
Most vendor ROI pages give you one big number and hide the assumptions. This one shows the arithmetic, separates what is checkable from what is projected, and tells you which lines do not apply to your agency. Read it skeptically — it was written to survive that.
Two different numbers. Never conflate them.
Agencies hear savings and think of one thing. There are two, and they differ in both size and certainty. Treating them as one number is how a good product loses a credible buyer.
| What it is | Size | How defensible | |
|---|---|---|---|
| 1. Software cost replaced | Vendor invoices that stop arriving | $1.5K - $32K / yr | Very. It is arithmetic on invoices you already have. |
| 2. Revenue and time recovered | Denials prevented, downcoding avoided, clinician hours returned | $25K - $450K / yr | Directional. Depends on your starting baseline. |
The second number is an order of magnitude larger and an order of magnitude less certain. If a vendor quotes you the big one without telling you which assumptions produced it, the number is decoration.
Software cost replaced
The checkable one. These are the tools an agency typically pays for separately, at vendor list prices you can confirm without talking to us.
| Line item | Typical market price | AgenciesForge |
|---|---|---|
| EHR / charting | $59-150 per clinician / mo | Included, unlimited clinicians |
| Clearinghouse | $250-400 / mo | Included (Stedi partner) |
| ePrescribing | $50-100 / mo | Included (WENO, incl. EPCS) — enabled at onboarding |
| HIPAA fax | $40-80 / mo | Included (AI-routed) — enabled at onboarding |
| Telehealth | $35 / mo | Included (self-hosted LiveKit) |
| AI scribe | $90-150 / mo | Included, and it learns your operations |
| Outcomes platform | $50-100 / mo | Included |
| EVV vendor | $2-6 per client / mo | Included |
Worked example — a 30-clinician agency
By tier
| Tier | Typical stack | AgenciesForge | Avoided / yr |
|---|---|---|---|
| EHR Starter | ~$325 / mo | $199 / mo | ~$1,500 |
| EHR Professional | ~$820 / mo | $499 / mo | ~$3,800 |
| EHR Enterprise | ~$3,780 / mo | $1,099 / mo | ~$32,000 |
Per-seat pricing is what makes the gap widen over time: their bill grows with every hire and ours does not. Unlimited clinicians on every tier, no per-seat fees, ever.
Charting and documentation
The workflow every clinician feels daily, and the largest recoverable number in the model. One dictation produces the note and everything that should have followed it.
| Today | With AgenciesForge | |
|---|---|---|
| Time per note | 30-45 min | 10-15 min |
| How it is produced | Typed after hours, from memory | Ambient scribe drafts SOAP, BIRP, DAP or GIRP from the encounter |
| Grounding | Blank template | Chart diagnoses, active ISP goals, recent notes |
| What else one dictation produces | Nothing | Safety plan when suicidal ideation surfaces, claim hint with CPT, ICD-10 and units, portal homework, diagnoses to add, goal-progress notes |
Say the hours out loud even when you skip the dollars. Two and a half hours a week back, per clinician, survives skepticism that a dollar figure does not — and it is the number that slows clinician turnover, which nobody puts in a P&L.
Billing and claims
Where the money already earned goes missing. Each lever below is a separate mechanism, not one effect counted four times.
| P&L lever | Without | With AgenciesForge | Annual impact |
|---|---|---|---|
| Denial rate | 8-15% | 3-5% — Stedi 270/271 eligibility plus a 15-point AI pre-submit audit | +$50-90K |
| First-pass paid rate | 75-82% | 88-94% — auto-appeal engine clears routine denials without human review | +$40-80K |
| Underpayment recovery | Unnoticed | Payments below 80% of your own learned median flagged as appeal candidates | +$10-25K |
| Authorization leakage | ~1.5% of billed, unbillable | Auth-unit drawdown tracked against the authorization, not a spreadsheet | +$15-45K |
Why it compounds
Every denial the platform sees teaches it. A CareSource denial for H0036 without an HQ modifier becomes a pre-submit warning on every future claim to that payer. The fee-schedule learner reads your 835 remittances and builds your own observed per-unit payment, so the charge-amount audit compares against your median rather than a national average. Month twelve catches what month one could not.
OASIS
The highest-dollar-per-document workflow in the platform. OASIS drives PDGM payment, so a mis-scored assessment is a downcoded episode — and you find out a month later.
| Today | With AgenciesForge | |
|---|---|---|
| Assessment capture | 60-90 min, blank form, post-visit | 25-40 min, chart-aware conversational interview |
| Item coverage | Clinician memory of what is required | Every clinician-answerable item for the timepoint (~198 at SOC), read from the CMS v3.02.0 spec |
| Chart context | Re-looked-up per item | Diagnoses, meds, wounds, SDOH, prior OASIS scores pre-loaded per question |
| Version drift | Manual bulletin reading | Registry pins the current version; per-version item deltas answer whether a change applies to this patient |
| Pre-submission validation | Find out at iQIES rejection | Local validator mirrors VUT — blocking issues surfaced before upload |
Electronic visit verification
EVV is where a delivered, documented, legitimate visit still does not get paid.
| Today | With AgenciesForge | |
|---|---|---|
| Capture | Separate vendor app, separate login | In the app the caregiver already uses; GPS and NFC tag options |
| Claim linkage | Reconciled by hand, monthly | Billed-vs-EVV reconciliation built in — every claim tied to its visit |
| Exceptions | Found when the payer denies | Missed and late check-ins surface as action items the same day |
| Plan adherence | Caregiver reports what they remember | Assigned tasks diffed against what was captured |
| State submission | Separate vendor contract | Sandata / HHAeXchange / state-direct adapters, agency-native fallback |
The line that lands with owners: you did the visit, you documented the visit, you just cannot prove it in the format the state wants — so you do not get paid for it.
DD / IDD waiver operations
Typical bottom-line impact for a 500-individual provider: $200K to $500K a year. The largest components are compliance failures that are cheap to prevent and expensive to discover.
- MUI compliance and state-portal workflows prevent $1K to $50K per missed-window report and the Medicaid clawback that follows a late MUI. DODD, OPWDD, DDS, APD, ODP, HHSC IDD and BHDDA are all wired.
- DSP training and credential tracking prevents $50K to $200K recoupment events when a surveyor finds DSPs delivering billed services without current training. This is the single largest DD/IDD clawback pattern.
- Auth-unit drawdown keeps recurring schedules from ever exceeding remaining authorized units, which is the Medicaid-denial failure mode in waiver billing.
- HCBS Settings Rule documentation, required at waiver renewal, generated from your existing individual and setting data instead of assembled by hand.
Scheduling, front office, and compliance
| Lever | Without | With AgenciesForge | Annual impact |
|---|---|---|---|
| No-show rate | 20-25% | 12-18% — multi-touch reminders at 48h, 24h and 2h, portal, secure messaging, auto no-show detection | +$30-60K |
| Admin time | 10-15 hrs/wk on scheduling, reminders, reports | 2-3 hrs/wk — AI Forge agent plus derived action items | +$15-25K |
| Audit / survey prep | $15-30K consultant engagement | Self-served audit-bundle ZIP, surveyor companion, deficiency forecasting | +$15-25K |
The number that is already in the other sheets
AI Forge is the most-asked-about part of the platform, so it is worth being precise about where its value actually sits. It is not a line item alongside charting and billing — it is the mechanism inside them. The scribe that cuts note time, the audit that stops the denial, the interview that captures the OASIS, the exception that saves the EVV claim: those are all AI Forge, and every one of them is already counted above.
A vendor would normally give this its own headline number. Doing that here would double-count the same dollars twice and hand you the one inconsistency that discredits every other figure on the page. So the table below is a cross-reference, not an addition.
Where it is already booked
| What AI Forge does | Counted in | Which figure it is |
|---|---|---|
| Ambient scribe drafting the note | Sheet 2 — charting | The 2.5 hrs/clinician/week already assumes it |
| 15-point pre-submit claim audit | Sheet 3 — billing | The denial-rate drop from 8-15% to 3-5% is this |
| Payer-quirk and fee-schedule learning | Sheet 3 — billing | The compounding accuracy is this |
| Conversational OASIS interview | Sheet 4 — OASIS | The 60-90 min to 25-40 min capture time is this |
| Same-day EVV exception surfacing | Sheet 5 — EVV | The recovered mismatch claims are this |
| Derived action items and risk roster | Sheet 7 — front office | The 10-15 hrs/wk to 2-3 hrs/wk is this |
What is genuinely incremental
Three things the other sheets do not capture. Note that this is the smallest total on the page — which is what you should expect once the double-counting is removed, and a reasonable test of whether the rest of the numbers were built the same way.
| Lever | Without | With AI Forge | Annual impact |
|---|---|---|---|
| Lookup tax | 3-5 screens to answer what is Pat on, did the auth come through, when was the last note | Asked and answered in place, from the chart | +$25-30K |
| New-staff onboarding | 8-16 hrs of EHR training per hire, plus weeks at reduced speed | New staff ask the system how to do the thing, in context | +$4-8K |
| Supervisory oversight | Director pulls reports to find what is unsigned, aging, or at risk | Asked directly, answered from live data | +$8-12K |
| Standalone AI scribe licence | $90-150/mo as a separate vendor | Included | See Sheet 1 |
The honest summary: if you removed AI Forge and kept everything else, you would not lose $42K — you would lose most of Sheets 2, 3, 4, 5 and 7 as well, because it is what performs the work those sheets measure. The $42K is only what is left over once the value already attributed elsewhere is stripped out.
Roll-ups by segment
Same levers, summed for a whole agency of each type. Non-medical home care is deliberately scoped tighter than the rest — no OASIS, 485, PDGM or NOA — and saying so builds more credibility than padding it.
| Segment | Assumed size | Projected annual impact |
|---|---|---|
| Behavioral health outpatient | 500 active patients | $250-450K / yr |
| DD / IDD waiver provider | 500 individuals | $200-500K / yr |
| Home health | 500 patients | $200-400K / yr |
| Non-medical home care | 200 clients | $60-140K / yr |
Return by tier
| Tier | Capacity | Subscription | Projected impact |
|---|---|---|---|
| EHR Starter | 50 patients | $2,388 / yr | $25-45K / yr |
| EHR Professional | 200 patients | $5,988 / yr | $100-180K / yr |
| EHR Enterprise | 500 patients | $13,188 / yr | $250-450K / yr |
The number worth comparing against your subscription is not the monthly price — it is how much you recover per dollar spent. Your mileage depends on starting denial rate, payer mix, and clinician utilization, which is exactly why the calculator asks for yours.
The rules these numbers follow
Published because a number you can audit is worth more than a number you have to trust. If a line below does not describe your agency, remove it from your own total.
Still to validate.These magnitudes were calibrated against typical agency P&L structure, not measured across our own customer base. The moment a customer will go on record with a real before-and-after, that measured number replaces the corresponding projected line on this page. A real figure from a named agency beats every model here.
Before you use these numbers
Are these numbers measured or projected?
Projected. Every figure on this page was calibrated against typical agency P&L structure, not measured across our own customer base. We label it that way deliberately. The one exception is the software-cost-replaced table, which is arithmetic on vendor list prices you can verify yourself in about ten minutes. The moment a customer will go on record with a real before-and-after, that measured number replaces the corresponding projected line here.
Why do you separate software cost from revenue recovered?
Because they have very different sizes and very different credibility, and conflating them is how vendors lose the room. Software cost replaced is $1,500 to $32,000 a year and it is checkable arithmetic on invoices you already pay. Revenue and time recovered is $25,000 to $450,000 a year and it depends entirely on your starting denial rate, payer mix, and clinician utilization. The second number is an order of magnitude larger and an order of magnitude less certain. You should treat them differently, so we present them separately.
How is the denial-reduction number calculated?
We model prevention, not appeals. The figure is the share of denials caused by pre-submission errors — coding, eligibility, modifier, authorization or EVV mismatch — that an audit catches before the claim ships. We assume 40% of denied dollars are catchable that way. We deliberately do not claim an appeals-overturned rate, because that statistic does not survive fact-checking and costs credibility for every other number on the page.
What makes the billing accuracy improve over time?
Every denial the platform sees teaches it. A CareSource denial for H0036 without an HQ modifier becomes a pre-submit warning on every future claim to that payer. The fee-schedule learner reads your 835 remittances and builds your actual observed per-unit payment, so the charge-amount audit compares against your median rather than a national average. Month twelve catches errors month one could not.
Do these numbers apply to a small agency?
The percentages do; the dollar totals scale down with your volume. The roll-ups on this page assume 500 active patients because that is where the arithmetic is easiest to follow. A 50-patient agency on EHR Starter sees a projected $25,000 to $45,000 a year against a $2,388 subscription. The ratio holds better than the absolute number — which is why the calculator asks for your actual patient count instead of showing you ours.
What is the fastest way to get a number for my agency?
Run the savings check. It takes about two minutes and asks for your clinician count, monthly billings, and denial rate. It runs the same constants documented on this page against your inputs rather than our example agency, and it will tell you when an assumption does not apply to you — authorization leakage, for instance, is only counted if you track authorizations manually.
Now run it on your numbers.
Two minutes, no account. It asks for your clinician count, monthly billings, and denial rate, then applies the same constants documented above — and tells you which lines do not apply to you.