We build a custom AI Intake Agent that reads your forms, PDFs, and emails, then fills in your CRM or
practice-management system for you. Your team reviews and approves. The typing is gone.
US-basedHuman-in-the-loop reviewYou own the systemBAA available
The problem
Your team is re-typing what a computer can read.
Every new client or patient means the same 12 minutes of typing: the same fields, copied from a form
into your system, over and over. It's payroll you can't see, and it's where errors sneak in: a wrong
insurance ID, a transposed date, a denied claim you find out about weeks later. Multiply your
minutes-per-intake by your monthly volume. That's the number. That's what this fixes.
Hours of manual keying. A 6-operatory dental practice can lose ~24 hours a month re-typing new-patient forms.
Errors at the door. Manual keying runs ~1–4% wrong per field, so roughly 1 in 3 intakes carries an error into your records.
Onboarding sits in a queue. Packets pile up, and a new client can wait 1–3 days before they exist in your system.
It reads. It extracts. You approve. It fills your system.
A custom pipeline wired into the tools you already run, with a human on the last click, always.
1. It captures
Web forms, uploaded or scanned PDFs, emailed packets, or a guided intake chat: however your information arrives today.
2. It extracts
AI reads every field and runs it through your rules: formats, required fields, value lists, and duplicate checks. Each field comes back with a confidence score.
3. You review
Every record lands in a clean queue with the source document beside the extracted data. Unsure fields are highlighted. Glance, fix, approve.
Human on the last click. On approval, a clean record drops straight into your CRM,
practice-management software, or folders, logged, filed, and done. Nothing is written to your system
of record until a person confirms it. Fewer errors than manual typing, by design.
Cliniko, Jane, athenahealth, NexHealth as a Dentrix / Open Dental bridge.
New patients live in minutes
Legal & home services
Clio, MyCase for firms; Jobber, HousecallPro, ServiceTitan for home services.
Matter-open and job intake, populated
Accounting
QuickBooks Online and Xero: new-client onboarding packets and W-9s captured cleanly.
Fewer downstream filing corrections
Databases & files
Airtable, Notion, Google Sheets; filing into Google Drive, Dropbox, or SharePoint with a consistent naming convention.
Structured, searchable, filed
Legacy & bridged systems
Dentrix, Open Dental, Eaglesoft via vendor middleware, HL7 / import files, or a validated paste-ready record where no write API exists.
No hand-waving, scoped before you sign
OAuth where supported · scoped API keys in a secrets manager · US-region infrastructure · e-sign and notify hooks (DocuSign, Slack) optional
The payoff
What changes once it's running.
60–75% of keying time removed. Confirmed against your real numbers in month one.
Fewer errors, by design. The same validation rules run on every record.
New records live in minutes. Not days waiting in a queue.
Your data stays yours. US infrastructure, never used to train public models.
You own the system. A done-for-you build, not a per-seat subscription.
Your team, upgraded. They review instead of type, and get back to clients.
Worked example: a 6-operatory dental practice at ~120 new patients/month spends
~24 hours/month re-keying intake. Cutting that to a ~3-minute review
recovers ~18 hours/month, roughly $4,700/year in front-desk payroll,
back on the phones.
We model 60–75% of keying time removed and confirm it against your real intakes. We don't claim 100%, because a human still reviews.
Not a generic form. Not an off-the-shelf chatbot. A private system built for your fields and your tools.
You own it
A custom system wired into your tools, not another per-seat subscription. You keep the workflow, the configuration, and the data.
Human on the last click
Nothing hits your system of record until a person approves. Anything it can't read goes to a "needs manual" lane, never guessed in.
We tell you the real number
We measure extraction accuracy on your own sample intakes and show you before go-live. Bounded promises, no hype.
Built for practices and firms doing 20+ intakes a month. Typical targets we model with clients:
60–75% of keying time removed, 95%+ field accuracy on clean forms at go-live (90%+ on typed PDFs),
improving over time. Exact numbers are confirmed against your real intakes during month one.
Scoped to your channels, fields, and systems. You get a fixed build price after a 15-minute discovery call.
Essential build
1 intake channel (form or PDF), 1 destination system with a clean API, ~15–25 fields, standard validation, review queue, and training.
Fixed quote after discovery
Standard build
2 channels (e.g. form + PDF/email), 1–2 destination systems, ~25–40 fields, de-dup, folder filing, and an e-sign or notify hook.
Fixed quote after discovery
Advanced build
3+ channels including conversational or messy-scan OCR, a legacy/bridged system (e.g. Dentrix via NexHealth), multi-system write, and custom validation.
Fixed quote after discovery
Optimization retainer
Monitoring, error-rate tracking, integration maintenance, accuracy tuning, and new fields/forms as they change. Care, Growth, and Partner tiers.
Scoped after implementation · 6-mo minimum
Terms: 50% to start, 50% at go-live. Timeline is typically 3–5 weeks from signed SOW,
credentials, and sample intakes. The retainer runs the first six months, then month-to-month with 30-day notice. US clients only.
Not included (quoted separately): historical back-file migration · standing up a CRM/PM system you don't yet have · a new workflow for an unrelated process
Key's Touch is led by Key Scales, not an account manager
who hands you off to a junior team. When you book your Discovery Call, you're talking to the person who maps
your fields, scopes the exact write path into your systems, and stands behind the build. The conviction is
simple: you shouldn't have to choose between moving fast and doing it safely. So we do both: the latest
capability, the strongest governance, in the same engagement.
1st-Place AI Solution ArchitectPMP CertifiedAI & Data GovernanceFull-Stack Engineer
On clean, structured forms we target 95%+ field-level accuracy at go-live; on typed PDFs, 90%+. Handwritten and low-quality scans are lower and more variable, so we measure accuracy on your actual sample intakes and give you the real number before go-live, not a marketing number. Because a human reviews every record, the accuracy that reaches your system of record is your reviewer's accuracy, which is higher than manual typing.
Is there really a human checking everything?
Yes. By default, nothing is written to your system until a person approves it. The reviewer sees the original document next to the extracted fields, with uncertain fields highlighted, so review takes seconds. You can later choose to auto-fill certain low-risk, high-confidence fields, but that's off by default and you control where the line is.
What happens when it can't read something?
It never guesses into your system. Anything it can't confidently extract (a bad scan, an unfamiliar form, a garbled field) is routed to a "needs manual" lane in the review queue, clearly flagged. Your fallback is always a clean manual entry, never a silent bad record.
Where does our data go? Is it secure? Are you HIPAA-compatible?
Your data stays on US-based infrastructure, encrypted in transit and at rest, with scoped access limited to your workflow. We can sign a BAA for HIPAA-covered practices and a DPA where needed. Your intake data is not used to train any public AI model. We tell you exactly which systems the agent touches, in writing, before we build.
Who owns the system when it's built?
You do. This is a done-for-you build, not a subscription you rent. You own the workflow, the configuration, and your data. The monthly retainer keeps it monitored, maintained, and improving, but the system is yours. If you ever left, you'd keep the working system and the records it produced.
We use Dentrix / Open Dental / a legacy system with no real API. Can you still do this?
Often, yes, and we're straight with you about how. For systems without a clean public write API, we bridge through vendor middleware (like NexHealth for dental), through a supported import file or HL7 format, or, where a system genuinely can't be written to programmatically, the agent hands your reviewer a validated, paste-ready record so they still skip all the reading and re-keying. We confirm the exact write path during discovery and put it in the proposal.
How long until it's running, and how much work is it on our side?
Typically 3–5 weeks from signed agreement, your system credentials, and a set of sample intakes. Your lift: give us the sample intakes, connect your systems, and have one person available for a training session and light review during the first weeks. We do the building.
What does it cost?
Every build is scoped to your channels, fields, and intake volume after a short discovery. You get a fixed, phased quote with the payback math, based on your real intake volume, before you decide. The $1,500 Money Leak Assessment credits 100% toward the build. The optimization retainer is scoped after implementation (six-month minimum, then month-to-month).
Book a Free 15-Minute Discovery Call
Send us 10 sample intakes. We'll show you exactly what it saves.
A few quick questions so your call is sharp from minute one. Takes about 90 seconds.
Pick a time. That's it.
Choose a 15-minute slot. No pitch, no obligation. We'll map your fields, your systems, and give you a fixed build price and a real accuracy estimate.
15 minZoomYou'll talk to Key directly
Based on your answers, a full custom build may be more than you need right now. The payback window is
longest below ~15 intakes a month. For learning to automate on your own timeline, our sister brand
Key Builds AI has resources that will serve you
better today. When your volume grows, come back.
Send us 10 sample intakes. Keep the findings either way.
Book a free 15-minute Discovery Call. We'll map your fields, your systems, and give you a fixed build price and a real
accuracy estimate, measured on your own intakes, no obligation. US clients · no lock-in after
month six · you keep the system and the data.