Jigger is AI loss detection, live bar inventory, and a bar-intelligence agent grounded in your real data — under one roof. EDI ordering, invoice OCR, and a built-in recipe library. All on the phone you already own, replacing four tools and a Tuesday-morning spreadsheet with one.
"We're not replacing scales — we're replacing the Tuesday-morning spreadsheet. The phone you already own does the math the bar manager used to do by hand."
Most bars stitch four tools together: an inventory app, an analytics dashboard, supplier portals, and a spreadsheet. We replace all four — and the AI agent on top is what no one else has.
Five capture modes from a slider to a full video walk-by. Optional Bluetooth scale and FLIR thermal for opaque bottles. An AI photo-QA gate rejects bad photos before they hit the vision API. Live oz, par status, calibration, position memory — one ledger, every method.
A static threshold catches obvious leaks. A z-score model catches the quiet ones — bottles drifting two standard deviations off their own 30-day baseline. Per-bartender heatmaps attribute variance pro-rata by pour share. Competing-brand insights surface within-category anomalies. Every flag traces back to evidence, not vibes.
Ask in plain English. The agent calls nine live data tools, returns clean answers with inline tables, and suggests three follow-ups so the next question is one tap away.
Mix and match per bottle. The system reconciles whichever method any teammate uses. An AI photo-QA gate pre-screens every image — if a photo's too dark or too blurry, you get told to reshoot before the API ever sees it.
Tap-and-drag a digital level down the bottle. No AI. Always works, even with no signal.
Accuracy: 60–75%
Use when wifi is dead or bottle is unreadableOne bottle, one tap. AI estimates the fluid line, label-aware. Best for spot-checks on high-pour bottles.
Accuracy: 88–94% (clear glass)
Use for nightly spot-checks on hero brandsOne wide photo. The vision pipeline segments and reads each bottle in parallel.
Accuracy: 85–92% (clear glass)
Use to close out a single shelf in 12 secondsWalk slowly with the camera rolling. The pipeline extracts six keyframes per bottle and merges them with confidence-weighted best-of selection.
Accuracy: 92–96% (clear glass)
Use for full weekly inventory in 8 minutesHigher resolution and HDR than video, with re-shoot prompts driven by per-shot QA. Same merge pipeline as video walk-by, but each frame is intentional.
Accuracy: 93–97% (clear glass)
Use when lighting is rough but you have a momentFor dark amber, foil-wrapped, or velvet-bagged bottles where vision can't see in. Drop-on, drop-off — no permanent hardware. Manual fallback when paired devices time out.
Accuracy: ~99% (any bottle)
Pro tier · optionalA phone-attached thermal camera reads the temperature gradient through opaque glass. Wraps up the catalog of bottles vision can't solve.
Accuracy: 94–98% (any bottle)
Pro tier · optionalEvery photo passes through a local quality check first. Garbage photos get rejected with a "reshoot — too dark" prompt before the vision API charges a token.
Saves: bad inferences + API spend
Standard on every tierOpen · capture · confirm. Every mode lands at the same review screen showing oz, confidence, and a re-shoot button. Bartenders learn it in one shift.
Median learning curve: ~6 minutes
Same UX every modeCounts feed the operating layer. The operating layer feeds everything else — par alerts, reorder forecasts, the chat agent, the manager dashboard.
Every bottle's live oz is computed continuously: the most recent counted level, minus POS-recorded pours since that count. Recipes that share an ingredient bottle decrement the same row. Par-status updates in real time.
One-tap filters: All, Below Par, Critical, Most Poured. The bartender on close sees the right list in two seconds. Quick Audit lets a manager type-in a count without opening a session.
Color-coded by par status — green / amber / red — so the floor reads it across a noisy bar.
Every shelf bottle stores reference photos, an empty-bottle weight, and a position vector. After three or more confirmed scans the bottle is marked trained — vision accuracy tightens, and the system can detect when something's been moved or substituted. Position memory means the next walk-by knows what should be where.
The dollar threshold catches the obvious. The statistical model catches the quiet drift. Bartender attribution gets you from "something's wrong" to "it's happening on these shifts."
Counted ounces minus POS pours, valued at retail per-oz. Flag if the gap exceeds $20 or 4 oz on any bottle, any night. Catches obvious shrink, theft, and over-pours. Easy to explain to a GM in 30 seconds.
Each bottle has its own 30-day rolling mean and standard deviation. If tonight's variance lands more than 2σ above that bottle's normal noise floor, it's flagged as an anomaly even when the dollar threshold doesn't trigger. Catches the slow leaks that hide inside the noise.
| Brand | Counted | POS pours | Variance | Variance $ | Z-score | |
|---|---|---|---|---|---|---|
| Patrón Silver | 28.1 oz | 21.1 oz | +7.00 oz | $76.78 | +3.4 | FLAGGED |
| Tito's Handmade Vodka | 32.7 oz | 27.1 oz | +5.60 oz | $36.47 | +2.7 | FLAGGED |
| Casamigos Reposado | 14.2 oz | 12.0 oz | +2.20 oz | $18.90 | +2.3 | ANOMALY |
| Maker's Mark | 17.0 oz | 16.7 oz | +0.34 oz | $3.06 | +0.3 | |
| Hendrick's Gin | 6.9 oz | 5.9 oz | +0.96 oz | $9.60 | +0.8 | |
| Bacardi Superior | 9.7 oz | 9.0 oz | +0.71 oz | $4.26 | +0.4 |
Variance is attributed to each bartender by share of pours during the count window. Three shifts on Patrón, three shifts of attribution. The heatmap surfaces who's consistently associated with flagged bottles — without making accusations.
The system auto-detects when one brand inside a category — say, two well tequilas — drifts further than its peers. Catches substitution, mis-rings, and "we ran out of X so we poured Y" without a manager noticing.
Nine read-only tools wired into the live ledger. Multi-turn. Grounded — it never invents a number, because every claim traces back to a tool call.
Why is Patrón flagged tonight?
Patrón Silver counted at 28.1 oz consumed against 21.1 oz of POS pours — a 7.0 oz gap ($76.78), which is +3.4σ above its 30-day baseline. Two of three bartenders on tonight share its pour log; the heatmap attributes $48 to the 8pm–close shift. Suggested: pull the POS journal for that shift and compare to the camera if you have one over the well.
What should I reorder by Friday?
Three bottles will hit critical par before Friday close at current 30-day burn rate, day-of-week adjusted: Tito's (1.2 days), Patrón Silver (2.4 days), Casamigos Reposado (3.1 days). Combined PO drafted at $487 across two suppliers. Want me to pre-fill the order modal?
Compare bartenders on tequila pours.
Bartender A · 142 tequila pours · $0.08 variance per pour · clean. Bartender B · 96 pours · $0.31 variance per pour · within tolerance. Bartender C · 78 pours · $0.84 variance per pour · flagged on Patrón and Casamigos. C's nights also see the highest within-category drift — Patrón shows 18% more variance vs Casamigos than the other two bartenders' shifts.
Generate today's brief.
Tonight: $149 variance, 2 flagged, 1 anomaly. Three below par. Patrón needs reorder. Bartender C's shifts are running 4× the pour-variance of A's. Annualized burn at this rate: $54,400 / year. Top action: review the 8pm–close POS journal for tequila rings.
Find me a margin-friendly tequila cocktail to feature.
Three options ranked by margin against current pour costs: Paloma (78% margin, $11 menu, $2.42 cost), Margarita Classic (74%), Tequila Sour (71%). Paloma also uses Patrón economically (1.5 oz vs Margarita's 2 oz) — featuring it might pull volume off the shift where Patrón is bleeding.
Has any open order not been received yet?
One open order from Tuesday: SUP-RNDC-1042, $1,284 across 18 line items, 2 days overdue. No invoice OCR record. Suggest calling the supplier or marking as a discrepancy.
Every bar starts with 30 recipes — the classics, Texas signatures, and shots — fully linked to your inventory bottles. Pour cost and margin compute from your real pricing. Because recipes link to bottles, variance can be computed per recipe, not just per bottle.
Build over ice in a tall glass. Squeeze fresh lime, drop the wedge in, top with Topo Chico. Stir gently once. Glass: highball. Ice: cubed.
Add syrup and bitters to a rocks glass. Add bourbon and one large cube. Stir 20 seconds. Express orange peel oils over the surface, drop in. Glass: rocks. Ice: 1 large cube.
Salt half the rim. Build over ice in a highball. Add tequila and lime juice, top with grapefruit soda. Stir gently. Glass: highball. Ice: cubed.
Salt half the rim. Add tequila, Cointreau, lime to a shaker with ice. Shake hard 12 seconds. Strain over fresh ice in a rocks glass. Glass: rocks. Ice: cubed.
Most "ordering" features generate an email with a CSV attached. We generate a real EDI 850 v4010 Purchase Order document — the same X12 format the big distributors actually accept — and dispatch it however the supplier needs.
Real X12 documents uploaded to supplier endpoints. The format big distributors actually accept.
Pre-filled HTML order with mailto fallback. Works with any supplier on email.
JSON payload to a supplier's webhook URL. For partners with their own API.
Deep-link into the supplier's portal with the order pre-staged in the clipboard.
Click "Reorder Below Par." The modal opens with bottles already split into one PO per supplier, each prefilled with quantities. Approve each PO independently. Multi-line bottles route to whichever supplier holds the contract.
The forecast uses your 30-day rolling consumption with day-of-week weighting (Saturdays don't pour like Tuesdays). Order quantities are pre-filled to cover the next seven days plus a configurable safety buffer. Override any line manually.
UPCs, costs, supplier IDs all real-format. The supplier's EDI translator parses this exactly the way it parses orders from any other licensed system.
The bartender on the dock takes one photo of the supplier's paper invoice. AI extracts the line items, fuzzy-matches them to the open PO, and surfaces any discrepancies for one-tap review.
1. Snap the invoice. 2. AI extracts supplier name, invoice number, line items, quantities, and prices. 3. The system fuzzy-matches each line to the open PO. 4. Discrepancies surface in a review pane. 5. Approve received quantities — inventory recomputes immediately.
qty_short · supplier delivered fewer than ordered.
qty_over · supplier delivered more than ordered.
price_changed · supplier's invoice price differs from the PO price.
substituted · supplier swapped one SKU for another.
missing · line was on the PO but isn't on the invoice.
Reports fire automatically off the close session. They live in a dedicated tab inside the bar app, and they also surface as a widget on the Manager Dashboard at admin.conciergepilot.co — the same place every other Concierge Pilot product reports into.
Every bottle counted tonight, every variance, every flag, every anomaly, attributed by bartender. Lands as a markdown report in the Reports tab and as a notification on the manager dashboard.
"Generate today's brief" returns a shaped 4-paragraph summary: variance, flagged, top action, annualized burn. Save it, share it, paste it into Slack — or read it in 20 seconds.
One row across the top of the dashboard: tonight's variance · flagged count · below par · open orders · top three variance bottles. The same data the chat agent uses, formatted for a glance.
Critical par · a bottle hits zero or below safety threshold.
Statistical anomaly · a bottle clears the z-score threshold.
Order sent · a PO has been dispatched to a supplier.
Rate-limited to one alert per kind+bottle every 30 minutes. Slack via webhook. In-app for the manager dashboard. No alert spam at close.
Install on iPad, iPhone, or any Android device. The shell loads from cache for instant open, then sync's the live ledger. Works on the bar's existing devices — no new hardware to buy, no MDM rollout, no IT ticket. The web app is the app.
We do not claim 100% accuracy and we never will. Calibration improves results after roughly two weeks of real use. Optional Pro tier hardware overrides phone-only modes on opaque bottles.
| Capture method | Clear-glass | Tinted glass | Dark amber | Foil / velvet |
|---|---|---|---|---|
| Manual slider | 60–75% | 60–75% | 60–75% | 60–75% |
| Single photo + AI | 88–94% | 80–88% | 40–60% | ~0% |
| Shelf scan | 85–92% | 78–85% | 40–60% | ~0% |
| Video walk-by (6 frames merged) | 92–96% | 86–93% | 50–65% | ~0% |
| Multi-photo (HDR) | 93–97% | 88–94% | 52–68% | ~0% |
| + Calibration (after 2 weeks) | 94–97% | 90–94% | 55–70% | ~0% |
| + Bluetooth scale (Pro) | overrides | overrides | 99% | 99% |
| + FLIR thermal (Pro) | overrides | overrides | 94–98% | 94–98% |
| Invoice OCR (printed, lit) | 92–96% | 92–96% | 92–96% | 92–96% |
WISK is the best-known AI bar inventory product. We compete with their Premium tier — same monthly price, no setup fee, more AI. Every row below was verified against WISK's own help docs, marketing site, G2 reviews, and Capterra reviews.
| Feature | Jigger Pro | WISK Premium |
|---|---|---|
| AI Intelligence — where the fight is won | ||
| AI photo level-reading (multimodal vision) | Yes | No · scale or manual tap |
| Multi-photo merge with median across N reads | Yes | No |
| Video walk-by (10-15s pan, 6-frame parallel extraction) | Yes | No |
| Photo QA gate (auto-rejects bad frames before AI spend) | Yes | No |
| Conversational AI dashboard (9-tool live-data agent) | Yes | No · charts only |
| Statistical anomaly z-score vs 30-day baseline | Yes | No |
| Vendor cost anomaly alerts with renegotiation drafts | Yes · with AI draft | Alerts only · no draft |
| AI Menu Engineering (push/kill recommendations) | Yes | Best/worst seller view only |
| Shift forecaster (per-shift demand prediction) | Yes | Day-level marketing claim |
| AI Bartender Coaching (private per-bartender notes) | Yes | No |
| New Bottle Gap Analyzer (category coverage AI) | Yes | No |
| Conversational reorder ("draft Republic order under $1500") | Yes | No |
| AI Cocktail Inventor (specials from overstocked bottles) | Yes | No |
| Photo-to-Menu (snap a competitor's menu, AI extracts cocktails) | Yes | No |
| Auto-Spotter (vision diff between two shelf scans) | Yes | No |
| Daily Brief auto-generation (morning summary) | Yes | No |
| Capture & Calibration | ||
| Shelf scan (one wide photo identifies every bottle) | Yes | No |
| Bluetooth scale support | Yes | Yes (BYO scale) |
| FLIR thermal phone-clip (reads through opaque/dark glass) | Yes | No |
| Per-bottle Digital Twin calibration (position memory + training graduation) | Yes | No |
| Manual slider fallback | Yes | Yes |
| Inventory & Operations | ||
| Live oz formula (real-time, no nightly cron) | Yes | Yes |
| POS pour subtraction | Yes · same database | Yes · 60+ POS via API bridge |
| POS integration depth | Native · same DB row | Wider · API bridge |
| Recipe + cocktail costing engine | Yes · margin auto | Yes |
| Invoice OCR + supplier-learned mappings | Yes · LEARNED/AI/NEW badges | Paid add-on |
| Real EDI 850 ordering (X12 v4010) | Yes | Limited |
| Reorder draft modal | Yes | Yes |
| Loss Detection | ||
| Variance flagging (counted-consumed vs POS) | Yes | Yes |
| Bartender heatmap with pro-rata $ attribution by name | Yes | No · item-level only |
| Recipe-vs-actual substitution detection | Yes | No |
| Annualized loss projection ($/yr if shrinkage continues) | Yes | No |
| Competing-brand insights (category gap suggestions) | Yes | No |
| Supplier cost trends (90-day) | Yes | Per-line cost only |
| TABC regulatory export (Texas alcohol board) | Yes | No · Canadian-HQ vendor |
| Roles & Platform | ||
| Role gating with $ masking for bartender / viewer | Yes | Yes |
| PWA install + offline app shell | Yes · iOS + Android | Mobile app |
| Realtime live indicator (Supabase realtime) | Yes | Yes |
| Bundled with full restaurant ops platform | Yes · Concierge Pilot Tier 3 | Standalone only |
| Cost & Commitment | ||
| Monthly subscription | $499 / mo | $499 / mo |
| Setup fee | $0 | $1,500 |
| Annual contract required | No · month-to-month | Likely · gated |
| Free trial | 14 days · no card | Demo only |
| Year 1 total cost | $5,988 | $7,488 |
| Year 1 savings with Jigger Pro · $1,500 (the entire WISK setup fee) | ||
Sources verified 2026-04-30 against wisk.ai, help.wisk.ai, G2 product page, Capterra reviews, and recent product demos. We update this table when WISK ships new features. If a row is wrong, write hello@conciergepilot.co — we'll verify and correct.
WISK is the AI fight. Here's where everyone else lives.
| Jigger | Partender | Backbar | BarVision | Sculpture / Bevinco | MarginEdge + Freepour | |
|---|---|---|---|---|---|---|
| Approach | AI vision + scale + thermal | Manual tap on photo | Manual + Bluetooth scale | RFID smart spouts | Humans count weekly | Smart-scale platform |
| AI level reading | Yes | No · operator taps | No | n/a · pour metering | n/a | Scale-based |
| Hardware required | Phone (scale + thermal optional) | Phone | Phone (scale optional) | RFID spout per bottle | None (consultant brings) | Smart scale |
| Install time | Same day | Same day | Same day | 3-week install | Weekly recurring visits | Same day after scale ships |
| Annual contract | No | 14-day trial then m2m or annual | No | 1-year minimum | Quote-based | Month-to-month default |
| Setup fee | $0 | $0 | $0 | Hardware bundled | Quote-based | Onboarding gated |
| Typical monthly | $499 | $199 annual / $299 monthly | $79-$129/mo | $250-$300/mo | $250 self / $800 full service | $500/mo |
Most bars upgrade to Enterprise inside three months — multi-location and supplier integrations are where the platform earns its keep.
Single-location bar. Up to 100 bottles. AI vision included.
Same monthly as WISK Premium. Zero setup fee. 11 AI capabilities they don't have.
Multi-location operators, hospitality groups, supplier-direct integrations.
Direct answers. No fluff.