> ## Documentation Index > Fetch the complete documentation index at: https://help.1nspect.app/llms.txt > Use this file to discover all available pages before exploring further. # IT Tokens — The AI Currency # IT Tokens — The AI Currency **IT (Inspection Tokens)** are the metered currency for every AI operation on 1nspecT. Every AI call — narrative generation, vision analysis, schema extraction, polish, SPO synthesis — deducts IT from your tenant's balance. When the balance hits zero, AI calls are blocked until you top up; non-AI operations keep working. This article is the canonical reference: what costs what, where the balance lives, how to top up, and how to economize. *** ## Why a metered currency? AI operations call cloud LLMs (Google Gemini, Vertex AI) that cost real money per request. The IT token system means: * **You pay only for what you use** — no flat AI subscription baked into the platform price * **You can plan AI spend** — a typical inspection uses 30–80 IT depending on Vision usage * **High-cost operations are gated** — Vision (8 IT) and Voice DNA (50 IT) pre-flight-check the balance * **Free operations stay free** — FREE Library chips, repair-request generation, and several utilities cost 0 IT The bucket is **tenant-wide**, not per-user. All inspectors share the balance. *** ## Where the balance lives The IT balance is visible only on the **web admin** — there is intentionally no balance display on the mobile app. To check: 1. Sign in to the web admin as Owner 2. Navigate to **Settings → Subscription & Tokens** 3. The current balance and recent usage history appear Mobile inspectors see token data only via the **Insufficient IT Tokens** error when a call is blocked. *** ## Complete operation cost reference The authoritative table. ### Inspector-facing operations (mobile) | Operation | Codename | Cost | Where | | ------------------------- | ------------------------- | -------- | -------------------------------------------- | | AI Generate (text-only) | `generate-narrative` | **2 IT** | Capture Screen → ✦ AI Generate (Vision OFF) | | Vision Analysis | `vision-analyze` | **8 IT** | Capture Screen → ✦ AI Generate (Vision ON) | | Library Search 🪄 | `semantic-search` | **1 IT** | Capture Screen → 🪄 button | | FREE Library chips | *(local)* | **0 IT** | Capture Screen — automatic on caption typing | | Save to Narrative Library | `provision-library-entry` | **1 IT** | Capture Screen → Save to Library | | System Info Tag (photo) | `extract-metadata-photo` | **3 IT** | Capture Screen → ℹ️ Sys Info (with label) | | System Info Tag (text) | `extract-metadata-text` | **1 IT** | Capture Screen → ℹ️ Sys Info (text-only) | ### Admin-facing operations (web) | Operation | Codename | Cost | Where | | --------------------------- | ------------------------ | --------- | ----------------------------------------------- | | AI Draft (per finding) | `generate-narrative` | **2 IT** | Report Workspace → AI Draft tab | | Polish Narrative | `polish-narrative` | **2 IT** | Inspection Reviewer → narrative editor → Polish | | Polish Report | `polish-report` | **10 IT** | Inspection Reviewer → Polish Report action | | System Performance Opinion | `generate-spo` | **10 IT** | Ratings Screen / Reviewer → SPO card | | Bundle Findings | `bundle-findings` | **2 IT** | Reviewer — automatic at finding consolidation | | Voice DNA — extract persona | `extract-persona` | **50 IT** | Onboarding — one-time per inspector | | Schema extraction (OCR) | `extract-structured-ocr` | **3 IT** | Templates → PDF upload → AI extract | ### Free operations (0 IT) | Operation | Codename | | ------------------------------ | ------------------------- | | Repair Request generation | `repair-request` | | Property Baseline | `property-baseline` | | PDF schema extraction (no OCR) | `extract-pdf-schema` | | Structured text extraction | `extract-structured-text` | *** ## Typical inspection cost | Activity | Per-instance | Typical count | Total IT | | ---------------------------- | ------------ | ------------------------ | ------------- | | Photos + ratings (no AI) | 0 | 30–80 findings | 0 | | FREE Library chip selections | 0 | 15–30 selections | 0 | | AI Generate (text-only) | 2 | 5–15 findings | 10–30 | | Vision Analysis | 8 | 0–5 findings (selective) | 0–40 | | Save to Library | 1 | 2–5 saves | 2–5 | | Sys Info on equipment | 3 | 2–4 labels | 6–12 | | Operator: AI Draft | 2 | 2–8 findings | 4–16 | | Operator: SPO per system | 10 | 5 systems | 50 | | **Total typical** | | | **80–155 IT** | A heavy-AI inspection (Vision everywhere) can run 200–300 IT. A light-AI inspection (mostly FREE chips) can run under 50 IT. *** ## How balance updates work * **Pre-flight check (Vision):** before sending the photo to the model, the backend verifies balance ≥ 8 IT * **Deduct on success:** most operations deduct **after** a successful AI call. Failed calls don't deduct. * **Atomic:** the deduction is part of the same Firestore transaction as the result * **Fire-and-forget for fast operations:** some 1-IT operations deduct asynchronously to avoid latency *** ## Topping up 1. Owner signs into web admin 2. **Settings → Subscription & Tokens** 3. **Purchase Top-Up** — choose a pack size 4. **Stripe checkout** — pay with the platform's payment processor 5. **Balance updates within seconds** Pack sizes typically range from small (\~500 IT) to bulk (\~25,000 IT). Larger packs have a lower per-IT cost. *** ## Subscription vs top-ups | Mode | How it works | When | | ----------------------- | ---------------------------------------------------------- | ------------------------------------------- | | **Subscription bundle** | Your monthly subscription includes a baseline IT allotment | High-volume tenants (50+ inspections/month) | | **Top-up packs** | Bought ad-hoc | Low-volume or seasonal tenants | Most tenants run both. Unused subscription IT typically doesn't roll over — review your plan terms. *** ## Economy tips * **Build the narrative library.** Every saved narrative becomes a future FREE chip. After 3–6 months of library investment, FREE chips cover 60–80% of common findings. * **Use Vision selectively.** Save it for findings where the photo really tells the story. * **Polish only once.** Each Polish Narrative call costs 2 IT — don't run repeatedly. * **SPO generation late.** Wait until all findings are in before generating SPOs. * **Sys Info only when age matters.** For obviously-new equipment, skip the AI call. *** ## When balance is exhausted | Operation tried | What happens | | --------------- | -------------------------------------------------------- | | Vision Analysis | Blocked with the *"Insufficient IT Tokens"* alert | | AI Generate | Falls back to a **template CAR narrative** (free) | | Library Search | Error toast; FREE chips continue working | | Sys Info | Error toast; informational fields can be filled manually | The **inspection itself never blocks** on IT. AI is pure accelerator; the platform fully functions without it. *** ## Audit trail Every IT deduction is logged on `aiUsage` for the inspection. Review per-inspection AI spend in [Inspection Details → History](/1nspect/admin-operator/inspection-details). Cross-inspection AI spend lives in [Reports → Inspector Productivity](/1nspect/admin-operator/reports-and-publishing). *** ## Related articles * [AI Features Guide](/1nspect/features/ai-features) — operations in detail * [Mobile Settings & Tokens](/1nspect/inspector/settings-and-tokens) — mobile view * [Company Settings](/1nspect/admin-operator/company-and-pricing/company-settings) — Subscription & Tokens lives here * [AI During Capture](/1nspect/inspector/ai-during-capture) — per-operation mechanics