AI Orchestrator

A Specialized and Secure AI Orchestrator for Swiss Financial Compliance

Project Summary

Attribute	Value
Application ID	133.672 IP-SBM
Funding Program	Innosuisse Innovation Project
Duration	24 months (January 2026 - December 2027)
Total Budget	CHF 591,240
Innosuisse Contribution	CHF 327,336 (55%)
Partners	FHGR (Research) + WeCanGroup SA (Implementation)

The Challenge

Swiss financial institutions manually process 50-100 page scanned compliance documents including KYC forms, regulatory filings, and compliance questionnaires. Current AI solutions achieve 95%+ accuracy on single-page forms but degrade to 70-80% on extended documents.

This creates:

Processing times of 2-3 weeks per document
High error rates requiring manual review
Regulatory risk from incomplete extraction
Inability to leverage cloud AI due to data sovereignty requirements

Our Solution

An on-premise AI orchestrator that combines:

Capability	Description
Document Digitization	OCR + LLM extraction from 50-100 page scanned documents
Dynamic Field Mapping	Zero-shot schema mapping to CRM systems without manual configuration
Document Pre-filling	Automated population of forms with variable structures
Data Sovereignty	Full FINMA, FADP, and GDPR compliance with on-premise deployment

Quantifiable Targets

Goal	Target	Validated At
Document accuracy	90% on 50-100 page docs	MS5 (M20)
Schema mapping	F1 > 85% zero-shot	MS4 (M16)
Hallucination reduction	40% vs baseline	MS3 (M12)
Processing time	< 2 hours (from 2-3 weeks)	MS5 (M20)
Deployments	3-5 Swiss institutions	MS5 (M20)
Technology readiness	TRL 3 to TRL 7	MS5 (M20)

Current Status (v1.9.0)

Metric	Value	Status
Project Phase	M01 Day 4	On Track
Execution Plans	24/24 months	Complete
Total Tasks	314	Complete
Deliverables	19/19 templates	Complete
Validation	0 errors, 0 warnings	Passed

Quick Links

Section	Description
Overview	Detailed project description and vision
Objectives	8 quantifiable goals (OBJ1-OBJ8)
Milestones	5 time-based milestones (MS1-MS5)
Work Packages	WP1-WP5 detailed breakdown
Deliverables	19 deliverables across 24 months
Partners	FHGR and WeCanGroup information
Architecture	Technical system design
Progress	Current project status
Project Inventory	Complete goals, metrics, specifications
Showcase Demo	End-to-end capability demonstration
Developer Setup	Environment setup and quick start
Onboarding	Start Here guide for new team members
Tools Reference	All 15 scripts and tools documented
Glossary	Domain and project terminology
Data Pipeline	PDF-to-wiki data flow documentation
API Reference	REST API documentation (template)
Project Goals	Quantifiable objectives with targets
Risk & Contingency	72 risks with mitigations

Presentations

Document	Description
Comprehensive Presentation	24-slide Beamer deck with all deliverables (v1.9.0)
Project Overview	18-slide overview with charts
Executive Summary	1-page stakeholder summary
Kickoff Agenda	January 15, 2026 meeting agenda

Project Charts (10 Visualizations)

Chart	Description
Project Timeline	24-month Gantt chart
System Architecture	Technical stack
Resource Allocation	FHGR/Wecan hours
Objectives Overview	OBJ1-OBJ8 mapping
Data Flow	Processing pipeline
TRL Progression	TRL 3-7 roadmap
IP Strategy	Patent landscape
Revenue Model	Pricing tiers
Funding Structure	CHF 591,240 breakdown
Pivot Scenarios	Risk contingencies

Project Timeline

2026                                    2027
Jan    Apr    Jun    Dec    Apr    Aug    Dec
 |      |      |      |      |      |      |
 M1    M4     M6    M12    M16    M20    M24
 |      |      |      |      |      |      |
Start  MS1   MS2   MS3    MS4    MS5    End
       Foundation    Review  Match  Valid

Contact

Research Partner: FHGR - Fachhochschule Graubunden
Implementation Partner: WeCanGroup SA
Repository: github.com/wecangroup/wecan-innosuisse-ai

Funded by Innosuisse - Swiss Innovation Agency