Solution
Document automation
Some administrative work consists of classifying and copying data from documents into software. A machine can prepare regular cases; the sample, thresholds and exception queue decide how far.
The gains from automation
Cycle time is measured on the workflow
The pilot measures the time to prepare, check and correct, before and after. What you gain depends on the formats, the variability and the exceptions.
Entry errors become detectable
Extraction, format validation and reconciliation reduce some errors. A plausible but wrong value remains possible and needs a control matched to the risk.
The decision stays with a person
The machine prepares, sorts and proposes. It doesn’t approve a payment or reject a case: a confidence threshold decides what comes back to a person.
In practice
- Reading attachments, PDFs and scanned documents, including imperfect handwriting
- Extraction of the fields that matter: amounts, dates, references, parties, IBAN
- Automatic sorting by document type and routing to the right department
- Consistency checks before writing: does the amount match the purchase order, does the reference exist
- An explicit confidence threshold: below it a person decides, and their choice improves the setting
- Writing into your existing management software rather than into a new silo
Systems involved
- Shared mailboxes and watched folders
- Accounting and ERP over APIs or files
- Document management and archiving
- Multifunction scanners
- Line-of-business software, through its exchange interface
Service lineArtificial intelligence →
The incoming flows
The same work, against each sector’s own constraints. Every card opens the full sector.
Banking and insurance
Client onboarding and KYC files
Identity papers, registers and forms extracted and cross-checked at intake volume, with confidence thresholds and a decision trail a reviewer can replay.
Healthcare and life sciences
Document extraction and classification
Structuring of reports, forms and notes, with human review of anything that falls outside the expected shape.
Logistics and supply chain
Transport documents
Reading dockets, freight invoices and declarations, with consistency checks before anything is written.
Manufacturing
Structured quality records
Paper records and forms taken in automatically, with out-of-tolerance values escalated to a person.
Retail and e-commerce
Catalogue enrichment
Missing attributes filled from supplier sheets, with review before publication.
How it runs
One document type
We start with the highest-volume, most regular flow. One type handled end to end beats five handled halfway.
Measurement
The correct-extraction rate is measured on your real documents, awkward cases included, not on a clean sample.
Production
Alongside the manual process at first, long enough to check the two agree.
Extension
A second type, then a third, reusing what is already there.
To go deeper
The EU AI Act now applies: four questions for a Swiss company
A Swiss address does not settle whether the EU AI Act applies. Start with the company’s role, the EU connection, the effect on people and the evidence needed before production.
An AI agent in practice: its work, and where it stops
Bank reconciliation in an accounting practice, incoming mail in a municipality, quotations in an SME, research monitoring in an institute: realistic AI agents, with their guardrails.
Dependencies and next steps
Test the fit: Document automation
Describe the context, constraints and decision you need to make. The first conversation qualifies scope, boundaries and the next useful step.
Describe the situation