AI for home-care operators: win the audit before it starts.
You run a home-care or home-health agency. Your caregivers are good at care. The paperwork is what keeps you up at night: activity notes on a fixed clock, certifications and CPR cards that expire on their own schedule, 90-day service reviews, EVV rules, and the audit that tests all of it at once. This guide is what I build for owners like you, in the order that pays.
Fix documentation first: let caregivers speak their notes and have the log fill itself. Then make it impossible to schedule a caregiver whose certification has lapsed. Then wire EVV through your state's existing aggregator, automate review reminders and signatures, and put a watcher on your state's rule changes. PHI handling and a Business Associate Agreement get scoped in writing before anything touches patient data. Every build is judged on one number your auditor already checks.
Where agencies lose the audit before anyone shows up.
None of these rules are hard alone. Together, across dozens of caregivers and hundreds of visits a week, they become a second job nobody was hired for.
- Activity notes on a clock. State rules can require caregiver activity notes at short, fixed intervals during every visit. A caregiver doing real care is not writing. The gap shows up in the audit.
- Expirations nobody is watching. Certifications, CPR cards, TB tests, background checks. Each one expires on its own date. One lapsed caregiver on the schedule is a finding.
- 90-day service reviews. Due dates, the right people in the room, signatures collected. Miss the window and the file is incomplete.
- EVV. Electronic visit verification is required for Medicaid-funded home care, and in several states the data flows through an aggregator such as Sandata. Rejected visits mean delayed or lost payment.
- Rules that change quietly. State bulletins and rule updates land in an inbox nobody has time to read. You find out at the audit.
Six builds, in the order that pays.
| # | Build | What it does | The number it moves |
|---|---|---|---|
| 1 | Caregiver app with voice notes | The caregiver talks. The app turns it into the log entries your state requires, stamped with time, and flags a gap before the shift ends instead of after the audit. | Notes on time |
| 2 | Certification alerts that block scheduling | Every certification and CPR date tracked. Reminders go out weeks ahead. If a date lapses, the scheduler refuses the shift until it is fixed. | Expired certifications on the schedule: zero |
| 3 | Service review reminders and signatures | 90-day reviews scheduled automatically, the right people invited, signatures collected on a phone, the file closed without a chase. | Reviews completed on time |
| 4 | EVV integration | Visit data sent through your state's existing aggregator interface, such as Sandata, so visits verify without someone fixing them by hand. | Visits rejected or held |
| 5 | State-rule watcher | Watches your state's bulletins and rule updates, then tells you in plain English what changed and what it means for your agency. You and your counsel make the call. | Surprises at the audit |
| 6 | Physician and referral outreach | Keeps the doctors, discharge planners and case managers who send you clients warm, with timely updates and follow-ups that actually go out. | Referrals per month |
Build on it or rebuild.
A lot of agency owners already paid a developer to build some of this. Then the developer went quiet, the app half-works, and nobody can say what it would take to finish. That money is gone either way. The only question is what the next dollar should do.
So I never start by writing code. I start with a rescue audit:
- Who owns what. The code, the hosting account, the domain, the app store listings. If you do not hold the keys, that is fixed first.
- What actually works. Tested on a real phone, with a real caregiver workflow, not a demo account.
- Where patient data lives. How it is stored, who can see it, and what has to change before real clients go in.
- What finishing costs. Measured against what a clean rebuild costs.
The audit ends with a written call: build on what you have, or rebuild. Either way you decide with facts, not with sunk cost.
Patient data gets scoped first.
I do not claim HIPAA compliance on a web page, and you should be careful with anyone who does. Compliance lives in how a specific system is built, hosted and run. Here is how I work instead:
- PHI and a BAA, in writing, at the start. How protected health information is handled, and the Business Associate Agreement, are scoped in writing at the start of every engagement. Before any system touches patient data.
- Care decisions stay human. The system writes down what the caregiver says. It does not decide care.
- The watcher flags. You decide. The rule watcher explains changes in plain English. Interpreting the rule for your agency stays with you and your counsel.
- Approved channels only. EVV runs through the interfaces your state and its aggregator publish. No scraping, no workarounds an auditor would question.
Where agencies waste the money.
Skip these
- A website chatbot before the documentation problem is solved. Your audit risk is not on your website.
- Another all-in-one platform your caregivers will not open. If it is harder than paper, they will use paper.
- Any vendor that says "HIPAA compliant" and cannot show you the BAA and where the data is stored.
- Pouring more money into a stalled build before someone independent checks what is actually there.
An operator in your corner, not another vendor.
I'm the Fractional Chief of AI for businesses that know they're behind. I take you from watching AI happen to running on it in 90 days. Two decades in enterprise tech: SRE and cloud architecture at Google, machine learning at BMW of North America. I run production AI systems today, including a legal-intake voice system with 50,000 calls behind it. For agencies, the engagement runs the same way as everything I do: one KPI, 90 days, your team trained to run what gets built.
Direct rates are published at /consult. Every engagement starts with a free 30-minute diagnostic call.
The questions I get every week.
Straight answers, no hedging. If yours isn't here, ask me directly.
What should a home-care agency automate with AI first?
Caregiver documentation. When state rules require caregiver activity notes at short, fixed intervals during a visit, the gap between what happened on a visit and what got written down is where audits are lost. Micah Berkley starts home-care engagements with a caregiver app that turns spoken notes into the required log entries, because it protects revenue and audit standing at the same time.
Is an AI system for home care HIPAA compliant?
Micah Berkley does not make a blanket HIPAA claim about any system on a website, and owners should be wary of anyone who does. Compliance depends on how a specific system is built, hosted and operated. In every Micah Berkley engagement, the handling of protected health information and a Business Associate Agreement are scoped in writing at the start, before any system touches patient data.
Can AI connect to our state EVV system?
Where the state or its EVV aggregator publishes an approved integration path, yes. Sandata is the aggregator in several states, and aggregators like it publish interfaces for this purpose. Micah Berkley builds EVV connections only through those existing, approved channels, never through screen-scraping or workarounds that an auditor would question.
We paid a developer and the project stalled. Can it be saved?
Often, yes. Micah Berkley starts with a rescue audit: who owns the code and the accounts, what actually works today, how patient data is stored, and what it would take to finish. The audit ends with a written call to build on what exists or rebuild, so the owner decides with facts instead of sunk cost.
What does AI for a home-care agency cost?
It depends on scope, which is why every engagement starts with a free 30-minute diagnostic call. Micah Berkley publishes direct rates at micahberkley.com/consult ($150/hour individuals, $240/hour business), and larger builds are scoped in writing before any work begins.