This printout holds only the sections loaded so far. For the whole dossier, print _full-print.html in this folder or use D3_web.pdf.

Research-Use Peptide Sales

5 sections · 32 pages · about 17 min
The four components

Four parts. One runs the day.

The Triage COO
AtlasTx

AtlasTx

One ranked list of everything open, fires first. Atlas works from record codes and never sees a name, and a person approves everything it drafts.

  • Fire card for safety reports and held orders
  • Five sort orders and a daily brief
  • Blind Atlas: record codes, never names
The framework
LastStop

LastStop

Intake, cases, clinical setups, pharmacy, patient portal, billing, audit and integrations.

The media
Growth Engine

Growth Engine

Brand kits, style kits, finished ads in every size, and outreach drafts.

Survival
ClaimRight

ClaimRight / Compliance

Claim rules, disclosures, consent, state rules, license checks, approvals and the audit log.

Norvane01 / 32
LastStop · the framework
Pillar · the framework

Research-Use Peptide Sales, run on LastStop

A research-use store with an AI COO, an ad studio and a compliance desk built in.

  • Norvane is a fictional brand made for this deck. Its name, logo and vial labels are invented, and any resemblance to a real company is unintended.
  • Its three sample products carry invented brand codes (A7, Q3, T12) that name no real compound. Every vial label is set by code.
  • Products in a store like this are sold for research use only, not for human consumption. Nothing in this deck is a health claim.
  • Every screen runs on labelled SAMPLE data. No orders, customers, revenue or certifications are claimed.
  • The storefront connection (WooCommerce or Shopify) is in the build plan, not built.
Fictional brandSAMPLE data throughoutResearch use only
Sample data
Animation poster

Section 1 The AI 5 pages · about 3 min

Norvane02 / 32
AtlasTx · the Triage COO
Pillar · the Triage COO

AtlasTx, your Triage COO

Held orders, missing COAs, low stock and flagged claims in one ranked list.

  • The store dashboard opens on AtlasTx. It watches orders, carts, support, stock, lot paperwork, payments and flagged claims.
  • Sample queue: a held order, open support tickets, a lot missing its certificate of analysis (COA), a low-stock product, two flagged product-page claims and five abandoned carts.
  • Atlas ranks every open item inside the app, with an AI re-rank available.
  • Nothing runs until someone presses Run AtlasTx.
Fires firstCodes, never namesSAMPLE data
Sample dataLive in demo · sample dataAtlasTx · AI re-rank available
01-dashboard-atlastx.png
Norvane03 / 32
Animation poster
03AtlasTx

The run

Illustration of a runFires come out firstRecord codes only
Illustration · the run itself is live in demo on sample dataAtlasTx queue
  • What a run does: orders, carts, tickets, stock, lot paperwork, payments and compliance flags stream in; AtlasTx takes them in; the ranked list comes out one card at a time.
  • This is an illustration drawn in code, not a screen recording. The item count in it is illustrative.
  • In the product, the run is one button: Run AtlasTx.

Press Run. It pulls in the whole store, then hands back your day in order.

Norvane04 / 32
02-atlastx-queue-and-fire-card.png
04AtlasTx

The queue and the fire card

OPS COORDINATOR−0.5FTE$2,500/mo
Fire card on topFive sort ordersA person clears each item
Sample dataLive in demo · sample data · real once the store is connectedAtlasTx queueAI · briefing (from the data only)
  • Sample fire card: an email scheduled for 11:00 to 1,900 customers contains a health claim. It is stopped for review before anything else.
  • Then the ranked queue: held payments, a missing COA, low stock, flagged claims, carts.
  • Five sort orders to choose from.
  • AtlasTx lines up the fix; a person clears each item.

Fires first. A health claim in the 11:00 email outranks everything.

Norvane05 / 32
04-support-tickets.png
05AtlasTx

Support tickets

SUPPORT REP−0.5FTE$1,850/mo
Store topicsUrgency read on every ticketA person sends
Sample dataPreview build · sample dataAI · support triage (drafts)
  • Tickets are about orders, shipping, COAs and lots, returns and payments. No clinical wording.
  • AI drafts an urgency read and a suggested reply with the order in view; a person reads, edits and sends.
  • Nothing is emailed or texted from this screen.

Order, shipping, COA and lot questions, each with a drafted reply.

Norvane06 / 32
05-follow-ups.png
06AtlasTx

Follow-ups

Draft one, or draft allCarts, holds, restocksNothing is sent
Sample dataPreview build · draftsAI · follow-up drafts
  • Every abandoned cart, held order and waiting restock request is listed with a drafted message.
  • A person approves, edits or discards. No messaging provider is connected, so nothing is sent.

Abandoned carts, held orders and back-in-stock requests, each with a draft.

Section 2 The Platform 4 pages · about 2 min

Norvane07 / 32
LastStop · the framework
Pillar · the framework

LastStop for a store

Keep your store. LastStop runs the work around it.

  • WooCommerce or Shopify stays the storefront. LastStop is the operating, creative and compliance layer above it, with AtlasTx as the AI COO.
  • The store connector (orders, customers, products, stock, lot and COA fields) is in the build plan. Nothing is connected yet.
  • Until then, every store screen runs on labelled SAMPLE data.
Your storefront staysOperations, media, compliance on topConnector in the build plan
Store connector: in the build plan
17-integrations.png
Norvane08 / 32
09-ai-at-work-roster.png
08LastStop

AI at work, store mode

11 stepsDrafts, sorts, explainsA person approves
Sample dataPreview build · sample dataRoster read from the code
  • The roster: the AtlasTx queue, a daily briefing, support triage, follow-up drafts, four outreach bots (cart recovery, back-in-stock, win-back, new leads), an operator assistant, provenance and replay, and brand-voice customer messaging.
  • The briefing never adds a number that isn't in the data.
  • Provenance and replay records which AI step ran, the wording it was given and what it read, and can re-run it to compare.
  • The list is generated from the code, not written by hand.

Eleven AI steps, listed from the code. None sends anything on its own.

Norvane09 / 32
17-integrations.png
09LastStop

Integrations for a store

WooCommerce / Shopify: plannedProcessor: not contractedNothing connected yet
Planned
  • On the page, all planned: card processing (processor not yet contracted), ShipStation labels and tracking, 3PL fulfilment, email and SMS, and analytics.
  • The store connector itself is in the build plan.
  • Each card says what is built, planned or waiting on an agreement. None is shown as live.
  • Ad-account hand-off is in the build plan: an export pack first, a direct push later.

Store, payments, shipping and messaging, each with an honest status.

Norvane10 / 32
m01-dashboard-atlastx.png
10LastStop

On a phone

390-pixel phone widthSAMPLE label stays visibleApprove taps still required
Sample dataLive in demo · sample dataAtlasTx
  • The dashboard, AtlasTx, the Ad Studio, follow-ups and support fit a 390-pixel phone screen.
  • SAMPLE DATA stays visible on the phone.
  • Anything that sends or changes a record still needs an approve tap.
  • Perfectly wrapped to fit your phone. iOS and Android apps are planned, for you and your patients.

The whole console fits a phone.

Section 3 Growth Engine 9 pages · about 5 min

Norvane11 / 32
Growth Engine · the media
Pillar · the media

Growth Engine

Brand kit in, finished research-use ads out, every line checked.

  • Brand Absorber reads the store's site into a brand kit; Style Absorber borrows a look without the identity.
  • Ad Studio builds finished ads for 31 research-use purposes in 7 formats, in the store's locked Seasons.
  • Outreach bots draft cart recovery, back-in-stock, win-back and first-touch messages. Nothing sends without a person.
  • The research-use rules run on every ad before a person sees it.
Brand kit inAds in 7 formats outResearch-use line on every ad
Ad Studio and Brand Absorber live · Seasons in a preview buildAI · brand absorberAI · image ad makerResearch-use check
Sample ads
Sample ads for a fictional brand.Nine Seasons, three variants each. Made outside the demo workspace and labelled SAMPLE. None has run as an ad.
Hover to pause. Click an ad to open it.
Norvane12 / 32
  1. 1Read the page
  2. 2Keep the kit DNA
  3. 3Ads in that kit
The page it read
Uploaded page · the fictional research-use store
The fictional store's home page Brand Absorber read: a copper bar, the logo, a serif headline and a copper button
Logo
Primary · #C4703A
Background · #16191C
Accent · #E2965C
Type · Georgia
Ink · #F4EFE7
The brand kit
Brand kitKit DNA
Primary#C4703A
Accent#E2965Clow confidence
Background#16191C
Ink#F4EFE7
Type
AaGeorgia · display
Inter · body
Logo
Sample ads in that kit
Sample ads

Built from the real run shown below: the store page Brand Absorber read, the colors it returned (it flagged the accent as low confidence), and Season samples for the same fictional brand. The sweep and the flying chips are this page’s animation.

Brand Absorber
Brand Absorber
12Growth Engine

Brand Absorber

BRAND DESIGNER−0.5FTE$2,600/mo
Reads the live sitePalette, fonts, logoYou approve the kit
Live in productAI vision + page read
  • Reads a site's palette, fonts and logo into the brand kit. An HTML upload works too.
  • The kit also holds the store's product photos and its pre-approved messages.
  • A person reviews and edits the kit before it is used.

Paste the store's address. Get the brand kit: colors, fonts, logo.

Norvane13 / 32
Style Absorber
Style Absorber
13Growth Engine

Style Absorber

Screenshot or page inStyle kit outNo logos, names or colors carried over
Built · labelled "New idea pending" in the productAI vision · style extractionIdentity-leak refusal
  • Rebranded? New style? Want to test a look, or learn from one that's working for someone else?
  • Upload a screenshot or give a page address. It extracts a style kit: layout, composition, type treatment, color treatment, photography, motion and copy approach.
  • It never carries identity: no logos, company or product names, copy text or exact color values. A check refuses any kit that leaks a hex color code.
  • If the page has no preview image, it fails honestly instead of guessing.

Borrow the look. Never the brand.

Norvane14 / 32
Sample1080 × 1080
NORVANE

Tracked shipping

From our bench
to your door.

See the catalog

For research use only. Not for human consumption.

Every size

One ad. Every size.

The same Season re-frames for each placement. The vial photo, the words, the button and the research-use line are placed by code, so they stay exact at every size.

Sample. The scene is the real Season 4 plate and the vial is the kit photo; the layout is set live in this page. Four of the seven formats.

14-ad-studio-brand-and-seasons.png
14Growth Engine

Ad Studio, research-use mode

AD DESIGNER−0.5FTE$2,600/moCOPYWRITER−0.5FTE$3,200/mo
31 purposes × 7 formatsUp to 6 product photosCode sets the required lines
Live in demoAI · image ad makerResearch-use check
  • Formats: feed 1:1, feed 4:5, Stories and Reels 9:16, display 16:9, email header, web hero, product card.
  • Copy comes from the store's pre-approved messages, for example "Every lot has its own certificate of analysis."
  • Every ad carries "For research use only. Not for human consumption.", set by code.
  • The rules block compound names and codes, benefit words, and dosing or human-use words on every ad.
  • The image model never draws the logo, the call to action or the research-use line.

31 research-use purposes, 7 formats, and the research-use line on every ad.

Norvane15 / 32
Sample ad
Seasons

One brand. Nine Seasons.

A Season is a saved ad style. Pick one or two Seasons; every later ad follows them.


One sample ad per Season, for the fictional brand. Labelled SAMPLE; none has run.

15-ad-studio-season-gallery.png
15Growth Engine

Seasons

Lock 1 or 2 SeasonsEvery ad traceable to its recipeMisspelled ads never ship
Preview buildAI image + code layoutText read-back
  • A Season is a saved, versioned ad recipe. Nine are active in the gallery.
  • The first ad arrives in three Seasons. "None of these?" opens the gallery of finished samples, each labelled SAMPLE.
  • Lock one or two Seasons as the house look; later ads rotate between them.
  • Every ad records the Season and version it came from, so a good ad can be traced and made again.
  • When the image model draws a headline, a text reader checks it: one retry on a mismatch, then set type. A misspelled ad is never shown as finished.

Pick a house look once. Every ad after that stays on it.

Norvane16 / 32
16Growth Engine

What the Seasons look like

Sample ads for a fictional brand9 Seasons × 3 variantsNo claims, no purity figures
Sample creativeAI image + code layout
  • Sample ads made outside this demo workspace: pictures from an image model, words, labels and required lines set by code. Labelled SAMPLE in the app's Season gallery.
  • Nine Seasons across, three variants down: a 3 × 9 grid.
  • Tone, placement and product alternate by rule, not at random, so the grid reads as one pattern.
  • Vials and their labels are placed by code. Every word the image model drew was read back letter by letter.
  • No people, no syringes, no health, benefit, dosing or human-use words, no purity figures. The research-use line is on every tile.
  • Sample creative for a fictional brand. Nothing here has run as a paid ad.

27 sample ads, 9 Seasons, one pattern you can see from across the room.

Norvane17 / 32
27b-asset-library_v2-filled-passed-checks.png
17Growth Engine

Asset Library

Uploaded · rescued · passedDownload as one packageStatus on every ad
Live in demo · browser session only
  • Three shelves: uploads, ads rescued by a research-use rewrite, and ads that passed.
  • Each ad keeps its status and its compliance verdict.
  • Download a set as one package, ready for an agency or an ad account.
  • Today the library lives in this browser session. A server-side library is not built yet.

Every ad that passed or was rescued, on one shelf, ready to download.

Norvane18 / 32
06-outreach-bots-1-and-2.png
18Growth Engine

Outreach bots 1 · Cart recovery and back-in-stock

LIFECYCLE MKTG−0.5FTE$3,300/mo
Cart recoveryBack-in-stockDrafts only
Sample dataPreview build · draftsAI · cart recovery botAI · back-in-stock bot
  • The cart recovery bot drafts email and text for customers who left a cart, and respects consent and opt-out.
  • The back-in-stock bot drafts notes for customers waiting on a sold-out product.
  • No messaging provider is connected, so nothing is sent.

Left a cart, or waiting on a restock? A drafted note, within consent.

Norvane19 / 32
07-outreach-bots-3-and-4.png
19Growth Engine

Outreach bots 2 · Win-back and new leads

Win-backNew leadsA person sends
Sample dataPreview build · draftsAI · win-back botAI · new-lead bot
  • The win-back bot drafts notes for customers who haven't ordered in a while.
  • The new-lead bot drafts a first touch for people who gave an email and never ordered.
  • A person sets the offer and approves every send.

Customers who went quiet, and sign-ups who never ordered, each get a draft.

Section 4 Compliance 11 pages · about 6 min

Norvane20 / 32
ClaimRight / Compliance · survival
Pillar · survival

Compliance: the section you don't skip

Two sample ads. One obvious, one with a single bad line. Both stopped.

  • Behind the stamps: two made-up sample ads, each failed by the research-use rules.
  • This section covers the scanner, the verdict, the research-use rewrite, the re-scan, the check with the AI unavailable, the explicit error, the site scan and the audit log.
  • The rules check five things: compound names, benefit claims, dosing, "low-profile" wording, and the research-use line.
  • No certification is claimed.
Research-use check firstThe AI can't overrule a findingEvery run audited
Sample dataPreview build · sample adsResearch-use check
Made-up sample ads
Made-up sample ads
Norvane21 / 32
19-scanner-checklist-and-findings.png
21ClaimRight / Compliance

The scanner reads the whole ad

CLAIMS REVIEW−0.5FTE$3,350/mo
Text in the image is readFive research-use checksExact words quoted
Sample dataPreview build · sample adResearch-use checkAI · reads text in images
  • The sample ad names a compound, promises a benefit and gives a daily dose, in the picture and in the caption.
  • The checklist is the research-use rule set: compound, benefit, dosing, low-profile, research-use line. Each is ticked or flagged from the real findings.
  • Each finding names the rule and quotes the exact words.
  • The rules decide. The AI reads the text in the image and cannot overrule a finding.

It reads the words in the picture too, and names the rule each one breaks.

Norvane22 / 32
20-scanner-verdict-fail.png
22ClaimRight / Compliance

The verdict

6 blocking · 1 warningA fix for each finding"Rescue this ad"
Sample dataPreview build · sample adResearch-use checkAI · review (cannot overrule)
  • Seven findings, six blocking and one warning, each with a research-use fix.
  • Then FAIL, and a button to rescue the ad.
  • The dosing line an earlier version of the scanner missed is caught here.
  • A second review agreed: FAIL.

Seven findings, six blocking. FAIL, with a way out.

Norvane23 / 32
23-rescue-rewrite-by-the-rules-after.png
23ClaimRight / Compliance

Fix a flagged ad

Before and afterNo compound, benefit or doseResearch-use line added
Sample dataPreview build · sample adAI · rewriteCheck before showing
  • Before: the failed ad and its findings. After: the rewrite of the same ad.
  • No compound name, no benefit, no dosing. The app adds the research-use line itself.
  • The note on screen: "Checked against the research-use rules before showing."
  • A person decides whether to use it.

The same ad, rewritten to the research-use rules, and checked before you see it.

Norvane24 / 32
Re-scan: FAILED to PASSED
Re-scan: FAILED to PASSED
24ClaimRight / Compliance

Re-scan: FAILED to PASSED

Same rules both timesRewrite composited on a pictureLabelled stand-in picture
Sample dataPreview build · sample adResearch-use checkAI · rewrite
  • The rescued ad is scanned again with the same rules: before FAILED, after PASSED.
  • The rewrite is set onto a new background by code.
  • The background in this capture is a labelled SAMPLE IMAGE stand-in, because the image generator was not called. The copy, its check and the verdict are real.

Before: FAILED. After: PASSED. Same scanner, same rules.

Norvane25 / 32
20-scanner-verdict-fail.png
25ClaimRight / Compliance

The check holds if the AI is down

AI service off for this testVerdict: FAILShown as proof, not the normal state
Sample dataPreview build · sample adResearch-use check
  • The same bad ad, with the AI service switched off for this test: the research-use check still fails it.
  • The check keeps working if the AI service is down.
  • This is shown to prove the point. It is not the normal running state.

If the AI service is down, the bad ad still fails.

Norvane26 / 32
21-rescue-explicit-error-edit-manually.png
26ClaimRight / Compliance

When nothing safe is left, it says so

Explicit errorNo filler resultEdit manually
Sample dataPreview build · sample adResearch-use check
  • With the AI unavailable and nothing safe left in the ad, the answer is: "Couldn't produce a research-use-safe rewrite — edit manually."
  • No fixed-up result is shown. A weak ad shown as finished would teach the operator the tool is weak.

No safe rewrite? You get an error, not a watered-down ad.

Norvane27 / 32
23-rescue-rewrite-by-the-rules-after.png
27ClaimRight / Compliance

One bad line, removed by the rules

One flagged lineGood lines keptLabelled as a rules rewrite
Sample dataPreview build · sample adResearch-use check
  • A second sample ad with one bad line (a weight-loss claim): two findings, FAIL.
  • Its rewrite while the AI is unavailable: the rules remove the bad clause, keep and join the good lines, and the app adds the research-use line.
  • Labelled on screen: "Rewritten by the research-use rules (AI unavailable)".

One bad line out, the good lines kept, the research-use line added.

Norvane28 / 32
25-site-scan-research-use-hits.png
28ClaimRight / Compliance

Site scan

COMPLIANCE MGR−0.5FTE$3,350/mo
Nine blockersNamed by ruleResearch-use line warning
Sample dataPreview build · sample page
  • Site scan of a made-up sample product page: production Fail.
  • Nine blockers named by rule (compound, benefit, dosing, low-profile), plus a warning for the missing research-use line.
  • The same rules as the ad scanner, so a page and an ad are held to one standard.

Paste a product page. Every blocker named by rule.

Norvane29 / 32
26-site-scan-clean-page-pass.png
29ClaimRight / Compliance

A clean page passes

Certificate of analysis linkedResearch-use line presentProduction pass
Sample dataPreview build · sample page
  • A clean sample page, with a certificate of analysis and the research-use line: production pass.
  • Passing is a rules result, not a certification.

A page with its COA and the research-use line passes.

Norvane30 / 32
28-audit-log-every-compliance-run.png
30ClaimRight / Compliance

Every run, in the audit log

Scanner, rewrite, site scanRefusals logged tooShort quoted words
Sample dataPreview build · sample dataAudit log
  • Each row: who ran it, the module, the verdict, the rules hit and a short quote of the words.
  • A refused rewrite is logged like any other run.
  • It is the evidence to show when a payment processor or ad platform reviews the store.

Every scan, rewrite and site scan leaves a row: who, what, verdict, rule.

Section 5 Jobs and Money Saved 2 pages · about 1 min

Norvane31 / 32
31Jobs replaced · totals

What it takes off the payroll

4.0 FTE of store work covered at about 500 orders a month.

RoleFTEBLS medianMonthlyAnnualLoadedBasis
−0.5
Operations coordinator (order exceptions)AtlasTx · SOC 43-5061
−0.5$59,650$2,500$29,825$42,610After store connection
−0.5
Customer service representativeAtlasTx · SOC 43-4051
−0.5$44,770$1,850$22,385$31,980After store connection
−0.5
Brand and identity designerGrowth Engine · SOC 27-1024
−0.5$62,960$2,600$31,480$44,980Built today
−0.5
Ad designer (production)Growth Engine · SOC 27-1024
−0.5$62,960$2,600$31,480$44,980Built today
−0.5
CopywriterGrowth Engine · SOC 27-3043
−0.5$76,910$3,200$38,455$54,940Built today
−0.5
Retention and lifecycle marketerGrowth Engine · SOC 13-1161
−0.5$78,760$3,300$39,380$56,260After store connection
−0.5
Research-use claims reviewerClaimRight / Compliance · SOC 13-1041
−0.5$80,730$3,350$40,365$57,670Built today
−0.5
Compliance manager (store-wide checks)ClaimRight / Compliance · SOC 13-1041
−0.5$80,730$3,350$40,365$57,670Built today
−4 FTEnon-clinical work covered (estimate)
$22,800per month, BLS median base wages
$273,735per year, base wages
$391,090per year, fully loaded (×1.43)

BLS Occupational Employment and Wage Statistics, May 2025 national medians (bls.gov/oes). Fully loaded uses the BLS Employer Costs for Employee Compensation ratio, $46.89 ÷ $32.82 = ×1.43 (bls.gov ECEC).

Norvane32 / 32
32Where it stands

Next: the suit of armor for your store

Storefront stays yoursConnector in the build planNothing connected today
  • Built: AtlasTx, the Growth Engine and the research-use compliance flows shown here, on labelled SAMPLE data.
  • In the build plan: the WooCommerce/Shopify connector; orders, customers and catalog screens; ad-account hand-off (export first, direct push later); a partner portal for affiliates; per-channel reporting; the social content engine.
  • Not claimed: orders, customers, revenue or certifications.

Keep your WooCommerce or Shopify store. Add the AI COO, the ad studio and the compliance desk on top.

You've seen every section. Now click around the real platform yourself.

Sample data. Advanced AI features are not active in the demo.

Hi, I’m Atlas. Ask me about this page.