AI Services catalogue
The 27 built-in AI extraction services, the documents they target, and how to use them.
An AI Service (also called a module) is a ready-made extraction schema for one class of business document. It already knows which fields to read and how to normalise them, so you never draw a zone — you pick the matching service, run it, and the fields come back filled. This page lists all 27 built-in services, what each one extracts, and where you select them.
The 27 built-in services
Scanix ships 27 services today: 26 document-type modules plus one "describe-it-in-plain-English" service, Natural Language Query. In Settings → AI Services they're grouped into Finance & Accounting, Procurement & Logistics, HR & People, Legal & Governance, Identity & Compliance, and Other.
The catalogue is fixed in the build — you can't add a new service type or rewrite a service's fields from the UI. You can turn whole services on and off, and override whether individual fields are required.
Screenshot
The Template Designer AI Services tab: the Module picker on the left with one service selected, the right-hand schema-preview panel showing that service's fields, types, and example values. — shot ai-capture-ai-services-01
Reference table
The Fields column is the number of fields each service extracts; Natural Language Query has no fixed schema, so it shows 0 until you describe the fields you want.
| Service | Fields | Target document |
|---|---|---|
| Invoices AI | 25 | Supplier invoices, utility bills, purchase orders — header, addresses, totals |
| Receipts AI | 11 | Travel & expense receipts — merchant, totals, category |
| Identity Documents AI | 22 | Passports, driver licences, national IDs, residence permits, US SSN cards — printed and machine-readable fields, cross-checked |
| Bank Statements AI | 15 | Bank statements — header, balances, totals |
| Contracts AI | 10 | Contracts — parties, dates, legal clauses |
| Natural Language Query | 0 | Any document — extracts the fields you describe in plain English (no fixed schema) |
| Delivery Notes AI | 11 | Delivery / goods-received notes |
| Purchase Orders AI | 14 | Purchase order headers (for three-way matching) |
| Packing Lists AI | 12 | Packing lists / packing slips |
| Bills of Lading AI | 15 | Ocean bills of lading and airway bills |
| Quotes & Estimates AI | 14 | Sales quotations and estimates |
| Credit Notes AI | 14 | Credit notes / credit memos |
| Expense Reports AI | 12 | Employee expense reports / claims |
| Timesheets AI | 11 | Employee / contractor timesheets |
| Payslips AI | 13 | Payslips / wage slips |
| Employment Contracts AI | 12 | Employment contracts / offer letters |
| NDAs AI | 11 | Non-disclosure / confidentiality agreements |
| Meeting Minutes AI | 12 | Meeting minutes headers |
| Business Reports AI | 12 | General business / financial / management reports |
| SOPs AI | 9 | Standard Operating Procedure controlled-document headers |
| Inventory Logs AI | 12 | Inventory / stock logs / count sheets |
| Tax Forms AI | 11 | Government tax forms (W-2, 1099, P60, VAT returns) |
| Utility Bills AI | 14 | Electricity, gas, water, telecom bills |
| Customs Declarations AI | 13 | Customs declarations (CN22 / CN23) |
| Certificates of Origin AI | 11 | Certificates of origin |
| Insurance Policies AI | 12 | Insurance policy headers |
| Medical Forms AI | 10 | Medical / healthcare forms (intake, referral, prescription headers) — treated as sensitive health data |
Natural Language Query
Natural Language Query has no fixed schema. Type a request such as "Find the customer email, the order number, and the shipping address" and the service builds the extraction shape at run time from your words — so you can capture fields from a document type that has no dedicated service.
No AI extraction
Every picker starts with a No AI extraction card — "Use only template zones + OCR. No tokens consumed." Choose it to capture with template zones and OCR alone, with no AI call. In a Job Group, No AI extraction acts as a clear-all that switches every service off.
What the fields come back as
Every extracted field is normalised to a canonical form after extraction.
| Field type | Returned as |
|---|---|
| Text | Trimmed free text |
| Date (ISO 8601) | YYYY-MM-DD (parses many input formats) |
| Currency | An amount plus an ISO 4217 currency code |
| Phone number | International format (E.164, +CC…) |
| Address | Structured: street, city, region, postal code, country |
| IBAN | Checksum-validated, spaced every 4 characters |
| SWIFT / BIC | 8 or 11 uppercase alphanumerics |
| Country code | ISO 3166-1 |
When you select a service, the right-hand schema-preview panel lists every field with its key, type, description, an example value, and a required / optional badge.
Several services also run cross-checks after extraction and surface the result in the field-verification panel — for example, invoice line items must sum to the total; a bank statement's opening balance plus credits minus debits must equal its closing balance; an ID document's printed fields must match its machine-readable zone.
How to select a service
The same catalogue appears in four places, each suited to a different job. Use the tabs below.
A template carries one service, which runs every time the template runs.
- Open a template in the Template Designer and select the AI Services tab.
- Under Module, click a service card — or No AI extraction to use only zones + OCR. The right panel previews the chosen service's field schema.
- (Optional) Click a field's required / optional badge to flip it. This is a per-template override; required fields drive the AI vision fallback when they come back empty.
- Set Provider routing and the Privacy posture (see below).
- Save the template.
Once a service is bound, the Viewer's primary button becomes Analyse with AI. To run more than one service on a single input, use a Job Group instead.
Selecting a service is only half the job
A service tells Scanix what to extract; it doesn't yet wire those values into your index fields. Open the Mapped Fields tab and point each index field at an AI: <Service> field — otherwise fields stay on Use OCR zone (unmapped) and the AI values won't fill them.
A Job Group can run several services at once. The group splits the incoming batch into document segments, and each segment is passed through every selected service.
- Open a Job Group in the Job Group Designer and select the AI Services tab.
- Tick every service this group should run. Real service cards carry a checkbox so it's clear several can be on at once.
- When more than one service is selected, the chips at the top of the right panel switch which selected service's schema is previewed.
- Set the group-wide Provider routing and Privacy (same controls as a template).
- Save the group.
Required overrides are read-only here
In a Job Group the required / optional badges are read-only — per-module required overrides aren't saved at the group level yet. Set required/optional on the template instead.
Run a service on a one-off import that has no template applied, straight from the Viewer.
- Open an imported document in the Viewer with no template applied.
- In the right sidebar, choose a service from the AI picker (sparkle icon).
- Click Analyse with AI — it OCRs and runs the chosen service in one pass. Each page is extracted as its own record, so a stack of independent documents (ID cards, receipts) each gets its own fields.
The ad-hoc picker is hidden whenever a template is applied — with a template, the template owns the AI decision.
Settings → AI Services is the global on/off control room. Open Settings, choose AI Services in the left navigation, and use the Document services card.
- Flip the switch beside any service to turn it on or off.
- Use the Search services… box to filter, or Enable all / Disable all.
- The card subtitle shows N of 27 active.
Settings save immediately — there's no Save button.
Off in Settings, still pickable in the designers
Turning a service off here does not remove it from the designer pickers — they always list all 27. A service switched off is enforced only when extraction runs: it's skipped for that document (logged as a per-document skip) rather than failing the whole batch.
Routing and privacy at a glance
Wherever you pick a service — on a template or in a Job Group — you also set how it runs.
| Control | Options | Default |
|---|---|---|
| Provider routing | Local first, cloud fallback / Local only / Cloud only / Pinned to one provider | Local first, cloud fallback |
| Cloud provider (cloud-only or pinned) | Anthropic (Claude) / OpenAI (GPT) | — (you must select one) |
| Privacy mode | Allow cloud egress / Local only | per template / per group |
Privacy mode: local only overrides any routing policy and hard-blocks the document from leaving the machine. When cloud calls are allowed, you must tick I acknowledge cloud egress before Scanix will make them.
Bring your own cloud
Cloud calls go straight from your machine to your own Anthropic or OpenAI account. Scanix never sees your key or your documents on that call and never bills for tokens. Bring-your-own cloud key is a plan entitlement — without it, the cloud options appear with a lock and an upgrade prompt; on-device extraction is always available.
Medical Forms hold sensitive data
Medical Forms AI values are treated as protected health information. If data residency matters, pair this service with Privacy mode: local only so nothing leaves the machine.
Good to know
Next steps
AI capture
The mental model: classify, extract, split, and learn — and how it stays private.
Analyse with AI
Run OCR and an AI service together in one click from the Viewer.
Build a job group
Run several services across a mixed batch and route each document to the right template.
Related
- AI capture — how AI Services fit alongside Smart Templates, splitting, and the learning loop.
- Analyse with AI — the one-click Viewer action that runs a service on the open document.
- Build a job group — select multiple services and route a mixed stack.