Audit-grade lead intelligence / local demo

Retroaktiv skattepotential, sourcad för granskning.

SkatteRevision analyserar historiska svenska bolagsdata, iXBRL-årsredovisningar, domänregler och myndighetssignaler för att hitta återbetalningscase som revisionsbyråer annars måste gräva fram manuellt.

Boka lokal demo PDF Back to showroom

What it is, and what it is not.

SkatteRevision is not a public SaaS product and not a replacement for authorized tax advice. It is a research and decision-support system that complements accountants, auditors, tax lawyers, and grant-backed innovation partners.

Lead intelligence

The system scans historical company signals to find where tax recovery work might be worth a human professional's time.

Evidence-first dossiers

Outputs are designed as sourced workpapers: what was found, what legal basis applies, what is missing, and why the next step is justified.

Private operational demo

The full system stays local because the data vault is large and valuable. Serious demos run directly from the development machine.

Positioning

We look backward. Most tools look forward.

Company databases usually sell current status. SkatteRevision works backward across fiscal years to find missed positions, changed interpretations, hidden operational signals, and under-reviewed deductions.

The four commercial recovery tracks.

Each track has different evidence requirements, risk gates, and professional handoff needs.

FoU tax relief

Finds companies where R&D-like work appears in historical reporting but may not have been converted into available payroll tax relief.

  • Development costs and intangible assets
  • Patent and technical staff signals
  • Six-year retroactive review window

Industrial energy tax

Identifies manufacturing and process industries where electricity use, permits, and case law may support recovery work.

  • SNI and process classification
  • SMP and environmental-report signals
  • Three-year retroactive review window

Leasing VAT

Reviews vehicle and operating lease patterns for possible VAT recovery signals that require contract and usage verification.

  • K3 lease disclosures
  • Vehicle registry enrichment
  • Human contract review handoff

Property VAT adjustment

Searches for investment and property-use patterns where the 10-year adjustment mechanism may matter.

  • Asset and investment history
  • Taxable versus exempt revenue split
  • Building permit and property signals

Risk and stop gates

Weak cases are stopped early. The system treats missing data as a sourcing task, not as permission to invent a number.

  • Bankruptcy and enforcement checks
  • Data-quality scoring
  • Red-team challenge before output

Partner handoff

The best output is not a final tax filing. It is a partner-ready dossier that lets a qualified advisor decide whether to proceed.

  • Evidence manifest
  • Missing-data checklist
  • Client interview prompts

Pipeline from data vault to dossier.

01 Ingest

Historical annual reports, iXBRL, registry snapshots, and local cache layers are normalized.

02 Index

SQLite and vector stores make historical company signals queryable before any expensive research.

03 Route

Hermes maps SNI and financial signals to the relevant specialist agents.

04 Challenge

Risk-agent and Kritikern look for reasons the case should be killed or escalated.

05 Dossier

Only high-quality cases become professional workpaper-style reports.

The system has real machinery behind the pitch.

The current codebase is a local, polyglot research platform rather than a clean product repository. That is acceptable at this stage because the sellable asset is the data, the lead intelligence, and the repeatable workflow.

Extraction layer

Python iXBRL parsers, corruption handling, Swedish taxonomy mapping, and bulk processing scripts.

Index layer

Local SQLite databases and vector indexes for fast historical lookup before web research.

Orchestration layer

Express API, HermesOrchestrator, Azure AI Foundry routing, rate limiting, and agent prompts.

Evidence layer

Core protocol, data-source hierarchy, insufficient-data handling, conflict checks, and PDF dossiers.

Operator layer

Dashboard panels for leads, telemetry, knowledge, scoring, reports, scripts, and local data inspection.

Why this is valuable to a partner.

For a revision or tax advisory firm

SkatteRevision can act as a second-look engine over existing and target-client portfolios. The firm keeps the client relationship and professional sign-off; TwistedStacks supplies the discovery and evidence-prep layer.

For an employer or innovation funder

The system demonstrates applied AI in a concrete, revenue-linked domain: data engineering, legal reasoning, local-first infrastructure, agent orchestration, and professional workflow design.

Demo format.

A serious demo should not upload the data vault. It should run locally and show the workflow against a sanitized or partner-approved company case.

1. Pick a case

Use a sanitized sample, a public company, or a partner-provided org number.

2. Show the chain

Walk from local data lookup to agent routing, risk challenge, missing-data flags, and evidence notes.

3. Discuss handoff

End with what a qualified advisor would need next: client interview, accounting files, contracts, or FOIA.