CarbonTrue Clear — Beverage CO₂ / Microbulk Invoice Truth & Recovery Desk

Executive summary

CarbonTrue Clear is a done-for-you beverage CO₂ / microbulk Invoice Truth Pack plus optional contingency Recovery Sprint for multi-location US restaurants, bars, QSR/c-store fountain programs, and hospitality portfolios (typically 5–40 sites) billed by NuCO2 (Linde), Airgas National Carbonation, or regional beverage-gas distributors.

AI extracts monthly/delivery invoices and the signed microbulk or cylinder schedule; deterministic rules flag open escalators vs. contract caps, fuel/energy surcharges vs. EIA diesel / published indexes, tank-rental creep, minimum-commitment overbill vs. metered usage, property-tax admin fees, and cylinder demurrage mismatches. A specialist releases the Truth Pack and, under LOA, pursues credits and forward freezes. Continuity Desk monitors subsequent months. Renewal Window Pack prepares operators 90–180 days before auto-renewal.

Final decision: Blueprint. Sells recovered/avoided beverage-gas spend dollars and a specialist-released evidence pack — not a customer-operated OCR dashboard, not legal counsel, not a gas-brokerage platform on day one.

Thesis

Hospitality operators already pay recurring opaque beverage-gas invoices (gas, tank rental, fuel/energy surcharges, admin fees, demurrage) under multi-year microbulk lock-ins with fill exclusivity and early-termination damages. Buyers already accept contingency cost-recovery for adjacent facility categories (utilities, waste, uniforms). An AI-native desk that productizes contract-vs-invoice truth for mid-market multi-unit foodservice — too small for enterprise procurement teams, too busy for AP line-item review — can recover credits and freeze forward rates with a human specialist as the only customer-facing chokepoint.

Discovery rationale

This run steered away from saturated regulatory-filing "engine" patterns into hospitality / local-service back-office invoice economics (Section 28). Twelve-plus fresh searches across UCO rebates, portable-toilet rental AP, document shredding, entrance-mat rental, unused airline credits, and beverage CO₂/microbulk showed the strongest CODE signal in customer-side beverage-gas invoice leakage: NuCO2 alone cites ~168,000 locations; federal court filings document automated escalators and fuel/energy surcharge disputes; a July 2026 trade teardown quantifies five-year overpay vs. cylinder exchange; Airgas fuel-surcharge litigation/settlement proves buyer willingness to contest gas surcharges.

Hard-diff checks confirmed this is not FillBlock (propane cylinder requal paperwork), not BreakTrue (ready-mix low-cylinder break investigation), not DetentionTrue / PortClock (ocean demurrage), not GarmentTrue / RefuseTrue / ElevTrue / FogTrue (different vendor ontologies), and not a customer-operated gas-procurement SaaS.

Candidate comparison

Rubric dimensions (1–5 each, summed×1): low trust burden, low task judgment, high intelligence threshold, regulation-as-moat, no physical labor, Sam Altman test, outcome pricing, gross margin, buyer urgency, competitive whitespace, novelty vs manifest, AI fit, active demand, budget proof, waitlist potential, MVP narrowness, distribution clarity, licensing feasibility, operational repeatability, speed to first revenue.

CODE validation

C — Consumer/buyer trend

Multi-unit hospitality continues to outsource beverage carbonation to industrial-gas majors (NuCO2/Linde, Airgas National Carbonation) via microbulk tanks and cylinder programs. Operators face rising layered fees while draft/fountain programs remain non-discretionary. Verified

O — Opportunity

The underserved job is not "buy cheaper gas" as a marketplace — it is proving, line-by-line, that invoices diverge from signed schedules (escalator, fuel/energy surcharge, tank rental, minimums, admin fees) and converting that proof into credits + forward freezes without the operator building an AP forensics team. Inferred from verified contract patterns + litigation + trade analysis.

D — Demand

• NuCO2 litigation (Townhouse Restaurant v. NuCO2) alleged automated 4–12% annual increases and fuel/energy surcharge classes — buyer pain filed in federal court. Verified
• Airgas fuel-surcharge class litigation/settlement shows customers contest gas delivery surcharges. Verified
• July 21, 2026 Advanced Mixology teardown teaches bars they overpay $10k–$15k over five years on typical microbulk vs cylinder. Verified
• Adjacent facility-cost auditors (P3, UtiliSave) already sell contingency recovery — buyers allocate budget to this class of outcome. Verified

E — Economic sizing

Beachhead: ~5,000–15,000 multi-unit US restaurant/bar/c-store operators with 5–40 locations on NuCO2/Airgas/regional microbulk or multi-site cylinder programs (inferred from NuCO2's 168k locations + multi-unit share uncertainty). If 8–12% of invoices contain recoverable leakage averaging $80–$250/location/month and a desk captures 25–40% of proven leakage as fees, a 200-portfolio book at $150–$400/location/year net revenue is a multi-million ARR path. Ranges are wide; treat as order-of-magnitude, not a census. Inferred

Rubric scorecard

Composite: 87/100 (sum of scores / 100). Clears evidence threshold.

Target buyer

Economic buyer: Multi-unit restaurant/bar CFO, COO, VP Ops, or franchisee-group controller managing 5–40 locations with NuCO2, Airgas National Carbonation, or regional beverage-gas contracts totaling ≥$1,500/month aggregate gas+rental+surcharge spend.

Day-to-day sponsor: Regional ops manager or AP lead who receives PDFs but does not reconcile to contract schedules.

Beachhead ICP: 8–30 unit QSR/fast-casual or multi-concept bar groups in Sun Belt metros with fountain + draft programs on microbulk, approaching renewal inside 12 months or showing visible surcharge creep.

Jobs-to-be-Done

• When invoices arrive monthly, help me know which lines are wrong without reading 40 PDFs.
• When a fuel/energy surcharge appears, show me whether it matches my contract and a published index.
• When escalators hit in January, prove whether the increase was noticed/capped as required.
• When renewal approaches, give me a negotiation pack so I am not locked into another 3–5 year overpay.
• Do it for me — do not hand me another dashboard.

Painful problem

Beverage-gas invoices stack base gas, tank/cylinder rental, fuel surcharges, energy surcharges, admin/property-tax processing fees, and minimum-commitment true-ups. Microbulk contracts commonly run 3–6 years with fill lockouts and liquidated damages. Operators rarely keep a working copy of the fee schedule next to AP. Litigation and trade analysis show systematic annual increases and surcharge disputes; AP pays on faith. Leakage compounds across locations.

The outcome we sell

Outcome: Specialist-released Beverage Gas Invoice Truth Pack plus recovered vendor credits and verified forward monthly savings — done-for-you. Customer does not operate OCR/AP software. Not legal counsel. Not a day-one gas brokerage.

• Truth Pack: contract clause map, exception ledger, dollarized findings, evidence exhibits, recommended ask.
• Recovery Sprint: LOA-authorized credit requests + forward freeze negotiation.
• Continuity Desk: monthly monitoring of subsequent invoices against the frozen schedule.
• Renewal Window Pack: 90–180 day pre-renewal RFP/renegotiation brief.

First one-feature MVP wedge

Evidence summary

• Verified NuCO2 serves ~168,000 hospitality locations (company About).
• Verified Townhouse Restaurant v. NuCO2 alleged automated escalators (≈4–12%) and fuel/energy surcharge classes; class certification denied; case dismissed with prejudice Dec 2021 — still primary evidence of buyer complaint patterns.
• Verified Airgas fuel-surcharge litigation/settlement shows surcharge contestability.
• Verified July 2026 trade analysis: typical bar microbulk vs cylinder 5-yr gap ~$10k–$15k; escalators 4–8%/yr common.
• Verified Contingency cost-recovery norms: P3 shares savings; utility auditors commonly 20–50% of recoveries.
• Inferred Mid-market multi-unit operators lack dedicated beverage-gas category managers.
• Unverified Exact % of NuCO2 invoices with material recoverable leakage nationwide — measure in pilots.

Claim table

Source-claim matrix

Market and demand evidence

NuCO2 positions itself as a leading beverage-gas provider to hospitality with ~168,000 locations across restaurants, c-stores, venues, and breweries. Airgas National Carbonation markets MicroBulk and Keep-It-Full programs nationally. Demand evidence is not "AI hype" — it is (1) court-documented surcharge/escalator disputes, (2) trade education telling operators they overpay, and (3) an existing contingency-audit budget class for facility invoices. Portable-toilet and UCO markets were sized as runner-ups ($3.3B porta rental; ~$2.8B cooking-oil recycling) but showed weaker DFY wedge clarity this run.

Active buyer conversations

• Federal complaints by restaurants/hotels against NuCO2 over price increases and surcharges (Townhouse / related FDUTPA theories).
• Airgas fuel-surcharge class claims by commercial gas customers.
• July 2026 bar-industry teardown walking owners through escalator, lockout, and termination-fee traps.
• Adjacent operator conversations in uniform/waste audit categories (P3 client narratives) prove willingness to authorize LOA-based recovery.

These are not waitlist signups; they are spend/complaint signals. Paid conversion still must be proven in pilots.

Competitive landscape

Competitor and budget validation

Existing budget: Operators already pay NuCO2/Airgas monthly. Adjacent budget for recovery already flows to contingency auditors (P3: share-of-savings; utility auditors: ~20–50% of recoveries). CarbonTrue redirects a slice of leaked beverage-gas spend into a pack + recovery fee — not a net-new "innovation tax."

Why incumbents are insufficient: Suppliers will not forensically challenge their own escalators. Broad auditors dilute attention across categories and often lack beverage-gas clause libraries. SaaS OCR still leaves negotiation on the operator.

Why we win: Narrow ontology (gas / tank rental / fuel / energy / admin / demurrage / minimums) + deterministic index checks + LOA negotiation productized for mid-market multi-unit hospitality.

Pricing evidence and proposed pricing

Benchmarks: utility/performance audits commonly take 20–50% of recoveries; P3 publicly describes contingency share-of-savings (often framed around multi-month shared savings windows) and 20–40% category savings findings.

Primary model is outcome/per-pack/per-portfolio — never hourly.

Regulatory and compliance considerations

• Commercial contract/invoice disputes — not a regulated professional filing service.
• Must not practice law: no court filings, no "you can void this contract" legal opinions; deliver commercial findings and operator-approved ask letters.
• LOA scope limited to obtaining billing records and requesting credits/adjustments.
• Handle customer invoices/PII with least-privilege storage, retention limits, and deletion on request.
• Contingency fees on commercial vendor credits are common in cost-recovery; avoid any fee structure that resembles debt collection against consumers (FDCPA out of scope — B2B vendor credits only).

Licensing boundary

AI-native advantage

AI changes unit economics by collapsing 12–24 months of multi-site PDFs into a structured ledger in minutes, mapping fee labels across NuCO2/Airgas dialects, and running deterministic validators (contract escalator formula vs. invoice; fuel surcharge vs. EIA diesel week; minimum commitment vs. delivered pounds). Humans concentrate on RELEASE judgment and vendor negotiation. As frontier models improve extraction/classification, pack COGS falls while the specialist chokepoint and frozen-schedule monitoring remain the defensible operating system.

Internal AI engine architecture

1. Intake — secure upload of contracts, invoices, site list, tank IDs; completeness checklist.
2. Normalization — OCR + vendor-specific parsers → controlled ontology (gas, rental, fuel, energy, admin, demurrage, minimum, tax).
3. Retrieval/knowledge — clause library, prior exception patterns, index feeds (EIA diesel, PPI if contracted).
4. AI workbench — clause extraction, label mapping, narrative draft of findings.
5. Deterministic rules — escalator math, surcharge formula, rental creep thresholds, minimum overbill, phantom delivery flags.
6. Human chokepoint — specialist RELEASE / HOLD / ESCALATE.
7. QA — second-pass sampling, dollar-tolerance checks, citation completeness.
8. Delivery — Truth Pack PDF + CSV + ask script; optional LOA negotiation tracker.
9. Learning loop — credited vs denied exceptions → rule/prompt updates.
10. Model-portability — prompts/rules/schemas versioned; swap LLM vendors without rewriting SOPs.

AI-vs-human operations pipeline

Dynasty translation layer

• Buyer: Multi-unit hospitality controller paying opaque CO₂ invoices wants proven credits and a cleaner forward rate.
• Service: DFY Truth Pack + optional Recovery Sprint; AI does extraction; specialist is trust interface.
• Workflow: Intake → normalize → retrieve clauses/indexes → AI draft → rules → human release → deliver → negotiate → monitor → renew.
• Tooling: Secure file intake, OCR, spreadsheet ledger, clause library, EIA feed, CRM, e-sign LOA — no custom platform day one.
• Sales: "We prove which beverage-gas lines violate your schedule and recover the dollars."
• Delivery: Manual LOA + PDF first; automate parsers per vendor later.
• Expansion: Continuity → Renewal Window → franchise packages → optional cylinder-vs-microbulk strategy briefs (still commercial, not legal).

Anti-duplication analysis

Checked against restored manifest (802 prior runs from tip 9dbeb3cf / FinalPay Clear) and repo *-blueprint.html filenames. No prior beverage CO₂/microbulk invoice-truth desk. Hard-diff:

• ≠ FillBlock / propane cylinder requal completeness
• ≠ BreakTrue ready-mix low-cylinder investigation
• ≠ DetentionTrue / PortClock / ChassisTrue ocean or chassis demurrage
• ≠ GarmentTrue uniforms/linen; ≠ RefuseTrue waste; ≠ ElevTrue elevator; ≠ FogTrue FOG manifests
• ≠ FinalPay Clear wage compliance

Same pattern family as invoice-truth desks, but different buyer trigger data (beverage-gas contracts), vendor ontology, and outcome dollars.

Anti-commoditization analysis

If general models let AP clerks paste PDFs into ChatGPT, they still lack: (1) vendor-specific ontology, (2) deterministic index validators with audit trails, (3) LOA negotiation playbooks, (4) continuity monitoring SOPs, (5) gold-standard credited exceptions. The company wins as an operating system for beverage-gas truth — not as a chat wrapper. If suppliers publish perfect self-serve dispute portals, Continuity + Renewal Window still monetize portfolio vigilance.

Service delivery workflow

1. Sales → Free Leak Scan intake
2. Paid Truth Pack engagement + document checklist
3. Secure intake + completeness gates
4. AI extract + rules + specialist release
5. Customer review call (20–30 min)
6. Optional Recovery Sprint LOA
7. Credit posting verification
8. Offer Continuity / Renewal Window

Operations as product

• Structured intake checklist (contract pages, rate exhibits, 12–24 mo invoices, site/tank map)
• Required evidence list before RELEASE
• Automated completeness scoring
• Exception queues by type (escalator, fuel, energy, rental, minimum, admin, demurrage)
• Reviewer assignment by vendor dialect
• Confidence scores + audit trail of every variance formula
• Versioned clause library + gold packs
• Root-cause tags on denied credits → rule updates

No-holes quality engine

No pack ships without: (a) cited contract clause for each material exception, (b) invoice exhibit page refs, (c) formula worksheet, (d) specialist signature block, (e) QA sample of ≥10% of dollarized exceptions or all exceptions if <$5k. Red-team monthly: inject synthetic overbill lines and confirm detection.

What the human expert actually does

Minimum viable offer

First paid offer: 12-month Invoice Truth Pack for one multi-unit portfolio ≤15 locations on NuCO2 or Airgas — $2,200–$4,500 founding price — delivered in ≤10 business days from complete intake. Recovery Sprint optional after pack acceptance.

Fulfillment process

First 3 customers (manual): Google Drive/Dropbox intake → Claude/GPT extraction into spreadsheet → specialist builds pack in Google Docs/PDF → Zoom review → LOA via HelloSign → email negotiation with supplier AR. Automate parsers after 5 packs. Do not automate RELEASE.

Tools and systems

• Day one: shared drive, spreadsheet ledger, Doc templates, e-sign, CRM (HubSpot free/starter), EIA diesel page, email.
• Day 90: vendor parsers, exception DB, QA sampling sheet, index auto-pull.
• Year 1: light internal workbench; still no customer-facing co-pilot.

Human-in-the-loop quality control

Every material dollar finding requires specialist RELEASE. Dual-control for Recovery Sprint settlement acceptances >$5,000. Customer approves LOA and any settlement language. AI drafts never auto-send to suppliers.

Nonlinear scaling and unit economics

COGS stack: model inference, hosting/storage, specialist review minutes, QA sample, LOA e-sign, CRM, rework, sales follow-up. Target revenue/FTE >$250k by month 12 via Continuity attach.

Conversion assumptions (explicitly assumptions): Leak Scan→Pack 25–35%; Pack→Recovery 40–55%; Recovery→Continuity 50–65%; annual Continuity logo churn <20%.

Distribution proof table

Sales and outreach plan

Plain offer: "We compare your beverage-gas invoices to your signed schedule and recover the difference. Free 3-site Leak Scan. Pack if findings clear the bar." Pitch focuses on cash leakage and renewal traps — not AI. Demo is a redacted sample pack, not software.

Founder-led content plan

Teach escalators, fill lockouts, fuel vs energy surcharge math, tank rental creep, minimum-commitment traps, renewal notice windows, and how to read a NuCO2/Airgas invoice. No generic "AI for restaurants" content.

First 30 days of content

1. 10 educational posts: escalator anatomy; fuel surcharge vs EIA; energy surcharge vs PPI; tank rental creep; minimum pounds; property-tax admin fees; cylinder demurrage; lockout clause plain-English; early termination math; multi-site AP blind spots.
2. 3 diagnostic teardowns: redacted invoice with flagged lines; contract exhibit vs January increase; microbulk vs cylinder TCO worksheet.
3. 2 lead magnets: "Beverage Gas Invoice Leak Checklist"; "Renewal Window 90-Day Countdown Sheet."
4. 1 live review: office-hours Leak Scan for 5 operators.
5. 1 outbound diagnosis template: personalized 5-line email citing their supplier + visible surcharge pattern ask.

Lead magnet and waitlist plan

Lead magnet: Free Leak Scan (≤5 sites / 3 months) → PDF teaser + waitlist for full Pack. Captures supplier name, location count, monthly spend band, renewal month. Trust via redacted sample math. Sales-ready when spend ≥$1,500/mo aggregate and ≥1 material exception pattern or renewal ≤12 months.

Warm GTM plan

Activate: prior restaurant contacts, CPAs serving multi-unit groups, waste/uniform audit partners, franchise consultants. Offer complimentary Leak Scans to their clients in exchange for warm intros (no mandatory rev-share at start).

Targeted outbound plan

Weekly list of 40 perfect-fit portfolios. Message leads with a diagnosis hypothesis ("January escalator + fuel line pattern on NuCO2-style invoices") and Leak Scan CTA — not a generic demo ask.

Answer-engine/search visibility plan

Publish citation-backed pages answering: "NuCO2 fuel surcharge explained"; "microbulk vs cylinder cost"; "how to prepare for beverage gas contract renewal." Structure with FAQs, tables, and primary-source links for AI answer engines.

Pilot design and early-demand-trap mitigation

Pilot cap: 8 portfolios. Incentive: founding Pack pricing + Recovery at 25%/20%. Feedback via structured exception outcomes. Custom legal research is out of scope — commercial packs only. Do not scale by hiring negotiators to paper over missing rules.

Early-access feedback flywheel

Every credited/denied exception tags root cause → updates ontology, rules, ask scripts. Product feedback = missing fee labels, false positives, index gaps. Custom work = one-off litigation support (refuse / refer out).

Build-before-scale checkpoints

• After 5 pilots: harden intake evidence requirements + QA sampling.
• After 10: harden SOPs, exception queues, reviewer checklists, delivery templates.
• After 20: pause new logos until COGS, rework, escalation, cycle time measured; no human-throwing at gaps.

7/30/90-day launch plans

7 days

Clause library v0 for NuCO2 + Airgas; sample pack; Leak Scan landing page; LOA template; 20 outbound; 3 warm intros.

30 days

10 educational posts live; 15 Leak Scans; 3 paid Packs; first Recovery Sprint in flight; parser v0 for top fee labels.

90 days

12 Packs completed; Continuity on ≥4 logos; measured win rate; decide go/no-go on hiring second specialist vs automation investment.

Metrics and KPIs

• Leak Scans / week; Pack close rate; $ leakage found / Pack; $ credits posted; forward savings verified; Continuity attach; cycle time intake→RELEASE; rework %; specialist utilization; gross margin; NPS/referral.

Risks and mitigations

• Supplier stonewalls: mitigate with precise exhibits + customer escalation path + Renewal Window leverage.
• False positives: dual formula check + specialist RELEASE.
• UPL creep: scripts reviewed; no litigation marketing.
• Contingency concentration: keep Pack flat-fee revenue ≥40% of mix.
• Manifest stub races: restore from last good tip before append (this run restored 802 from 9dbeb3cf).

Exhaustive risk register

What could kill this

• Suppliers move to fully transparent self-serve dispute portals that post credits automatically (reduces Recovery attach).
• Inability to obtain contracts at scale (intake failure).
• Regulatory reinterpretation treating contingency vendor-recovery as unauthorized practice in some states (monitor; keep commercial posture).
• If pilots show <10% of portfolios have material recoverable leakage, Pack-only economics may not clear CAC — then no-go expansion.

Go/no-go reasoning

GO — Blueprint. Clears evidence threshold: clear buyer, painful opaque invoices, court + trade demand signals, existing contingency budget class, narrow MVP, remote fulfillment, 50%+ margin path, strong hard-diff vs prior manifest, licensing boundary manageable with LOA + non-legal posture. Runner-up UCOTrue is promising but previously deferred and harder on remote volume proof; porta/air-credit/mat/shred fail whitespace or evidence bars this run.

Final recommendation

Build CarbonTrue Clear as a service-first beverage CO₂/microbulk Invoice Truth Pack with optional Recovery Sprint and Continuity Desk. Launch on NuCO2/Airgas mid-market multi-unit hospitality. Cap pilots at 8. Measure leakage find rate, credit win rate, and specialist minutes before scaling. Do not build customer-facing software first.

Source list

1. NuCO2 — About (168,000 customer locations)
2. Townhouse Restaurant v. NuCO2 — Complaint (CourtListener)
3. Order denying class certification
4. Joint stipulation for dismissal with prejudice
5. ClassAction.org copy of complaint
6. Advanced Mixology — Why bars overpay for CO2 (2026-07-21)
7. Airgas National Carbonation
8. Airgas fuel surcharge settlement coverage
9. Airgas fuel surcharge complaint excerpt (CourtListener)
10. P3 Cost Analysts — contingency cost recovery
11. P3 uniform/linen auditing (adjacent pricing norms)
12. Performance-based utility audit fee structures
13. UtiliSave restaurant utility audits
14. EIA gasoline & diesel retail prices (index reference)
15. IBISWorld Cooking Oil Recycling (UCO runner-up)
16. USDA AMS National Weekly Ag Energy Round-Up
17. Grease Trap Locator — switching UCO collectors (runner-up)
18. IBISWorld Portable Toilet Rental (deferred candidate)
19. Navan unused ticket credits (AirCredit reject context)
20. ITILITE unused flight tickets guide (2026)