Judgment work
Diagnosis. Treatment decisions. Consent. Capacity. Risk assessment. AI does not touch this. Not with supervision, not with a good prompt, not ever. Nine of these are named explicitly in the handbook as hard red lines.
AI is the intern.You are always the attending.
You already have ChatGPT. What you don't have is a method, one that says which tasks to point it at, what may never go into it, and how to verify what comes back before your signature goes on it.
Why "The Attending": in every workflow here, the AI is the intern and
you are the attending. The name describes your role. Not a credential we are claiming.
One payment. No subscription. Twelve months of updates included, and the files stay yours whether you renew anything or not.
Clinician-built·
Safety-first·
Built for healthcare professionals·
Designed around human review and data minimization
Plain text, travels anywhere
Epic
Cerner / Oracle Health
athenahealth
eClinicalWorks
Meditech
any system with a text field
01, What it produces
The part every other AI product hides
Most tools show you a beautiful finished note. We show you the thing that matters: exactly where the model was forbidden from filling in a blank.
You dictate~40 seconds
"ok 62 year old guy back again three days of cough worse at night no fever he thinks, ex-smoker quit maybe 2015, chest sounds ok to me, sats were fine, started him on the inhaler, review a week, told him red flags"
Draft returned for your review~90 seconds to check & sign
S: 62 y/o male, re-presentation. 3 days productive cough, worse nocturnally. Subjectively afebrile. Ex-smoker, ceased approx. 2015. O: Chest auscultation: [NOT DOCUMENTED] Oxygen saturation: [VALUE] Temperature: [NOT DOCUMENTED] A: [CLINICIAN IMPRESSION] P: Inhaler commenced, [AGENT, DOSE] Review in 1 week. Red-flag safety-netting given.
Read the right-hand column again. You said "chest sounds ok to me", the draft did
not write "chest clear to auscultation bilaterally." It wrote
Not documented and made you say it yourself. That single rule is the
entire product. It is a hallucination defense and a cloned-note defense at the same time.
02, The premise
Not all of your work is the same kind of work
The reason most clinicians get AI wrong is that they treat their entire day as one undifferentiated task. It is three, and only two of them are negotiable.
Category 01
Category 02
Expression work
You already know the answer; you are turning it into a note, a letter, an appeal, a handout. The clinical thinking is finished. Only the typing remains. This is where the hours actually are.
Category 03
Friction work
Reformatting. Reading-level conversion. Summarizing your own meeting notes. Restructuring a rota email. No clinical content at all. Start here on day one, the risk is close to zero.
03, The method
Four boxes. One rule. That is the whole system.
Every task you are tempted to hand to a model goes into exactly one of four boxes before you type a single word. Ten seconds of sorting, before the risk exists.
Don't
- Diagnosis and differential generation
- Treatment and prescribing decisions
- Consent conversations
- Capacity and safeguarding assessments
Double-check
- Coding and documentation sufficiency
- Guideline location (then verify at source)
- Anything with a number in it
- Anything a payer will read
Draft
- Notes from your own dictation
- Referral and discharge letters
- Medical-necessity documentation
- Patient-facing handouts
Delegate
- Reformatting and restructuring
- Reading-level conversion
- Meeting and teaching notes
- Rota, admin and inbox scaffolding
The Second Signature Rule
No AI-generated content enters a record, reaches a patient, or goes to a payer until a licensed clinician has read every word and taken responsibility for it.
04, The half nobody markets
Documentation that is accurate by construction
This system was designed to stop AI inventing findings. It turns out that the same architecture also prevents the other well-known documentation problem: cloned, templated, copy-forward notes that read identically from one patient to the next.
Mechanism 01
No clinical facts ship in the box
Every template and phrase carries the same footer: nothing here contains clinical facts, doses or thresholds. All clinical substance is yours. There is no normal exam waiting to be pasted, because there is no exam in the file at all.
Mechanism 02
Wildcards force per-patient content
Phrases arrive as skeletons with visible stops. An unfilled wildcard is glaringly obvious in a signed note, which is the structural opposite of the identical-ROS bloat that triggers review.
Mechanism 03
Non-fabrication is written into the output
Not documented is not an error message. It is a positive attestation, inside the note, that a finding was not obtained. An accurate record of what was and was not done.
The rule that stops a model inventing a heart sound is the same rule that stops a template inventing one. Accuracy and safety turn out to be the same engineering problem.
If you already have an ambient scribe
This is not a competitor. It is the other 70%.
DAX, Freed, Abridge, Sunoh. They write the encounter note, and they write it well. They have never written a medical-necessity letter, an AMA discharge, a critical-care time attestation, a handoff, a complaint response, a parent handout at a sixth-grade reading level, or a policy for your practice.
If your department is mid-rollout, this governs everything the rollout does not reach. If your department will never fund one, nocturnists, residents, locums, community clinics: this is the entire stack.
If your workstation is locked down
Citrix, no clipboard, no personal browser. Three routes in.
Every phrase in this package is plain text. The largest is under 2 KB. Included is a page covering the three ways clinicians actually get them into a restricted session:
- Type it once into the SmartPhrase / AutoText editor inside the session, after that it lives in the EHR permanently.
- Route through an org-approved document store your session can already open.
- Ask informatics for a one-time macro import. The format is exactly what they expect.
05, What you download
Thirty files. Every number on this page is generated from them.
No "and much more." No teaser bundle. This is the manifest. The same one that ships inside the download as a packing slip you can check off, file by file.
| Component | What it is | Count |
|---|---|---|
| Handbook | 43 pages: the safety architecture, the four-box method, the nine red lines, the three-tier PHI firewall, and seven end-to-end workflows. Plus a 4-page addendum, When NOT To Use This. Chapter 4. The tool-tier and de-identification chapter: is the one to read before you use anything. | 2 PDFs |
| Prompt library | Production prompts across twelve clinical-administrative categories. Every single one carries a triage badge: Green 58 prompts, use as written · Amber, verify before use 42 prompts, check the output against source first. 127 pages as a PDF, and the identical library as markdown with the same badges, for pasting. | 100 prompts |
| EHR phrases | Plain-text note skeletons with visible wildcard stops, ED MDM, AMA discharge, critical-care time attestation, consult, handoff, medical-necessity, portal message and more. Written in Epic dialect; the .txt format pastes or retypes into Cerner AutoText, athena and eCW macros unchanged. | 8 phrases |
| Word templates | Documentation pack, patient-communication pack, patient-education pack, and medical-necessity documentation letters. Editable, unbranded, yours to adapt. | 4 packs |
| Excel tools | Time Recovery Audit (measure your own baseline before and after), Prompt Performance Tracker, and Medical Necessity Tracker. | 3 tools |
| Checklists & cards | Pre-send verification, PHI firewall pocket card, red-lines card, attestation checklist, and the thirty-day rollout sequence. | Included |
| Bonus systems | The Ninety-Minute Inbox (fifty replies → six reusable structures), the Medical Necessity Kit, and the Teaching Kit. | 3 systems |
| Safe Harbor checklist | A 3-page checklist of the 18 HIPAA Safe Harbor identifiers, with the mosaic-risk test. The file to read before you paste anything anywhere. The free browser tool on this site is built from it. Try that first, it costs nothing. | 1 PDF |
| Start here | A one-page orientation whose entire doctrine is "choose one workflow. Only one." Written for someone opening this after a shift, not before one. | 1 file |
Every number on this page, and every image above, is generated from the files you download. The five images are rendered straight from the shipped PDFs. Not mockups, not a designer's impression. Counts were verified file by file against the archive.
Delivered as a single ZIP in numbered folders. PDF, Word (.docx), Excel (.xlsx), Markdown and plain text. The markdown and .txt versions exist specifically so you can paste into any tool without formatting fighting you.
06, Before you buy
What this is, and what it deliberately is not
We would rather lose the sale here than have you discover it after the download. Read the right-hand column first.
Yes This is what you get
- A written safety architecture you could hand to your legal or governance team
- Note, letter and appeal structures that force per-patient content
- Plain-text phrases that survive an EHR migration in your pocket
- A method for sorting AI tasks by risk in under ten seconds
- Tools to measure your own time recovery instead of trusting ours
- Twelve months of updates, and the files remain yours permanently
No This is what it is not
- Not a medical device and not clinical decision support. It suggests no diagnosis and no treatment, by design
- Nothing you buy makes an AI tool HIPAA- or GDPR-compliant. Your organization's approved-tool policy always outranks this package
- Not software. There is no app, no login, no integration, no plugin, these are documents
- Not an ambient scribe and not a replacement for one
- Not a published time study. We have not run a controlled trial, so we do not quote a number as if we had, see below
- Not specialty-complete. There is no obstetric, pediatric, psychiatric or ICU module today
On evidence, stated plainly
This product has not had external peer review, and we are not going to imply otherwise. No panel of physicians has signed off on it. It was written by one practising clinician and stress-tested against the failure modes catalogd in The 47 Documentation Failure Modes, which you can read in full, free, before you spend anything. We have also not published a measured time saving, because we have not run a study that would justify one. The Time Recovery Audit is included precisely so you can measure your own baseline, run one workflow, and measure again, and get a refund if the number does not move.
07, Pricing
One payment. No subscription. Ever.
The ambient-scribe market runs $200-$400 per clinician per month. This is a one-time purchase with twelve months of updates, and the files are yours to keep permanently, including if you refund.
Founding price
Clinician Suite
For an individual clinician, attending, resident, locum, PA or NP.
$99
$149
One-time · 12 months of updates · 30-day refund, keep the files
- All thirty files, complete
- Handbook, 100 badged prompts, 8 EHR phrases
- 4 Word packs · 3 Excel tools · checklists & cards
- 3 bonus systems (Inbox, Medical Necessity, Teaching)
- Instant download, delivered in under a minute
- Personal license, unlimited personal clinical use
Visa · Mastercard · Amex · Apple Pay · Google Pay · PayPal
Practice Suite
For a group, department or practice deploying this across a team.
$299
One-time · up to 10 clinicians · 12 months of updates
- Ten seats of the Clinician Suite (delivered as a second download)
- Practice AI Governance Pack, four adoptable documents
- Editable AI-use policy (.docx) your governance team can adopt
- AI Governance Register + SOP template
- ROI Calculator and Time Recovery Audit (.xlsx)
- Credential verification available to institutional buyers
- Invoice and W-9 on request
Need more than 10 seats? Email us
Card checkout is being switched on
The store is in final setup. If you want it now, email support@theattending.net and we will send your link at the founding price the moment it is live, usually the same day.
In the meantime everything you need to judge the product is already free and needs no account: the standard, the 47 failure modes, and the de-identification tool.
Why $99 today. The price ladder, published in advance
$99, now
Founding price. The product is complete and audited; the published evidence is not there yet, and the price says so.
$149
When a measured time study is published on this site with its method attached.
$199
When the specialty modules ship, admission & cross-cover, OB, paeds, psych, ICU.
Buy at $99 and every update inside those twelve months arrives free, including the modules above. We are not running a countdown clock. The price moves when the work ships, not when a timer expires.
08, Who made this
09, Questions
The ones that decide it
Is this safe to use with patient data?
Chapter 4 is the honest answer, and it is longer than you want it to be. The short version: the package uses a three-tier tool classification, a de-identification protocol, and a mosaic-risk test for whether "de-identified" actually is.
Nothing you buy makes an AI tool compliant. If your organization has not approved a tool covered by a BAA or equivalent data-processing agreement, the answer for identifiable data is no. And this package says so in those words. Four of the seven workflows use no patient data at all, which is where most people should start.
Does it diagnose, or suggest treatment?
No, deliberately, and by design. There is no diagnostic algorithm, no differential generator and no treatment recommendation anywhere in it. Those four things sit in the "Don't" box on page one and never leave it. This is an expression-and-documentation system, not a clinical one.
I'm on Cerner / athena / eClinicalWorks / a custom EHR. Is this Epic-only?
No. The phrases were written in Epic dialect because that is where they were built, but they
ship as plain .txt with visible wildcard stops, which is the most portable format in
this category. They go into Cerner AutoText, athena and eCW macros, or any text-expansion tool,
with a trivial change to the trigger syntax.
Everything else, letters, appeals, handouts, policies, the inbox system: is EHR-agnostic, because appeals and handouts live in Word and fax-land regardless of what your hospital bought.
How is this different from an ambient scribe?
A scribe listens to an encounter and writes that encounter's note. This covers everything a scribe has never touched: prior-authorization appeals, AMA discharges, critical-care time attestations, handoffs, complaint responses, referral letters, patient handouts at controlled reading levels, teaching material, and practice policy.
If you already have a scribe, these coexist, most of the reviewing physicians described them as complementary. If you will never be issued one, this is the whole stack.
Which AI tool do I need?
Any competent general-purpose assistant. The system is model-agnostic on purpose, prompts are written in plain language rather than tuned to one vendor's quirks, so it survives model versions changing under you. Use whatever your organization has approved.
I'm not technical. Is this over my head?
There is no code, no setup, no integration and no login. If you can copy and paste, you can use every file. The orientation page routes you to exactly one workflow to try first and explicitly tells you to ignore the other twenty-nine files until that one is habit.
What is the refund policy, precisely?
Thirty days, no questions, and you keep the files. We are not going to argue with a physician over a $99 purchase, and we cannot un-give you a download, so we do not pretend otherwise. Refunds are processed within three business days.
Email us before your bank. We will resolve it faster than a chargeback will. Full terms on the refund page.
Who reviewed it? Has it been peer reviewed?
Nobody has, and we are not going to pretend otherwise. This has not been through external peer review. It was written by one practising clinician and stress-tested against a catalog of forty-seven documented failure modes, which is published free so you can audit the reasoning yourself rather than trust a number of anonymous reviewers.
That costs us conversions and we are keeping it, because a product whose whole thesis is "verify before you sign" has no business asking you to take its own credentials on faith. The refund policy is the trial: thirty days, no questions, keep the files.
Is my specialty covered?
Partly, and here is the honest map. The documentation, letter, appeal, inbox and patient-education work is structural and applies across specialties. Dedicated modules for obstetrics, pediatrics, psychiatry, intensive care, and admission / cross-cover documentation do not exist yet - they are the next build, and they arrive free inside your twelve months of updates.
If your specialty is one of those and that gap decides it for you, wait for the module rather than buy today. We would rather you bought later than refunded sooner.
How is it delivered, and what payment methods work?
Instant download after checkout. A single ZIP, typically in your inbox within a minute. Payment is handled by Lemon Squeezy, who act as merchant of record: Visa, Mastercard, Amex, Apple Pay, Google Pay and PayPal, in your own currency where supported. We never see or store your card details, and any applicable VAT or sales tax is calculated and remitted by them.
Start with one workflow.
Not thirty files. One. Open the orientation page, pick the workflow that matches the task you dread most, and run it once tonight. If nothing changes in thirty days, take your money back and keep everything.
One payment · No subscription · 30-day refund, keep the files · Instant download