AI support and automation in the audit: what pruvia takes care of
pruvia is arguably the most innovative audit software in Switzerland: AI-supported, with audit documentation automated to the maximum so that the auditor can concentrate on the actual auditing. It carries out the limited statutory examination, the ordinary audit and special audits digitally, from the acceptance decision to the signed and delivered auditor’s report. The principle: you audit, pruvia documents. Every step creates its working papers as it goes, the AI suggests, and the auditor decides. This page shows step by step what happens automatically, where the AI helps and where it deliberately stays out.
As at 8 October 2026
The principle: suggestion, decision, evidence
Three rules apply in every function of pruvia. First: whatever can be derived from the data, pruvia calculates itself, deterministically and traceably at all times, for example leadsheets, key figures, materiality, capital situation or the 14 test patterns of the journal entry analysis. Second: where judgement is needed, the AI formulates a suggestion with reasoning; it becomes a working paper only through explicit adoption. Third: every adoption carries initials, time and origin. The audit file thus remains the work of the auditor, and the standards SER and SA-CH, which leave responsibility with the auditing person, are complied with.
What happens automatically in every audit step
| Audit step | Automatically by pruvia | Suggestion by the AI | Decision of the auditor |
|---|---|---|---|
| Data import | Read the trial balance from PDF, Excel or CSV, map accounts to the Swiss SME chart of accounts, create the financial year as its own file with previous year, build balance sheet and income statement. | Only for scanned PDFs without a text layer does the AI read the figures; for files with a text layer, pruvia works locally. | Confirm the column mapping, check accounts flagged as doubtful. |
| Acceptance and independence | Preconditions and independence questions as a checklist, declaration of acceptance of election as Word or PDF, engagement letter from a template with digital signing. | – | Give answers, approve acceptance, sign. |
| Planning and materiality | Benchmark with recommendation, materiality and performance materiality, key figure and variance analysis with automatically detected anomalies, depth of audit per area according to risk. | Market, environment and knowledge of the business as a text suggestion; for the ordinary audit, pointers on fraud under ISA-CH 240. | Assess risks, set materiality, approve planning. |
| Journal entry analysis | The entire journal against 14 fixed test patterns (journal entry testing): weekend and night-time entries, round amounts, splitting, Benford, manual entries, entries above materiality and more. | No AI, fixed rules: the same data always yields the same findings. | Assess each pattern as “risk” or “no risk”. |
| Audit areas and leadsheets | Leadsheets per audit area from the financial statements, audit programme according to the type of audit, evidence from the client portal lands directly in the right audit area, increased risk shows the extended procedures. | Audit procedure based on the stored documents, account reconciliation (does the balance match the document?), root-cause search when a trial balance does not balance. | Initial procedures, adopt or discard suggestions, close the area, review by the second person. |
| Inquiries and documents | Inquiry per audit area as a fillable PDF or via the client portal, answers flow into the audit field; request documents with deadline and automatic reminder; open items list from the audit areas, also to the client. | Wording of the questions on request. | Select questions, evaluate answers. |
| Ordinary audit | ICS audit with control inventory, bank confirmations under ISA-CH 505 with request letters per bank from the import, sample calculator (attribute and MUS), comprehensive report to the board of directors as Word. | Key controls from a process description. | Test controls, assess responses. |
| Meeting note | Record the meeting with the consent of all participants, transcription in Switzerland (AWS Zurich). | Draft with summary, statements, open points, tasks and assignment to the questions of the inquiry. | Only approval turns the note into a working paper. |
| Audit completion | Findings against materiality, capital situation under Art. 725 et seq. CO, preconditions for the report; closing meeting with client feedback via link and one-time code; representation letter with the financial statements as an enclosure. | – | Approve completion with initials, send the representation letter for signature. |
| Auditor’s report | Report from text modules in the EXPERTsuisse wording, preselection from the audit result (audit opinion, additional paragraphs on Art. 725 et seq. CO, notes), in German, French, Italian and English, in the audit firm’s design. | – | Check the wording, approve the report, sign with a qualified signature. |
| Signing and delivery | Digital signature via Skribble (QES for the auditor, SES for the client); after the signed representation letter, the report is delivered automatically, via link with one-time code instead of an e-mail attachment. | – | Sign. |
| Audit file and archive | Entire documentation as a PDF with evidence or ZIP with signed documents, archive versions, permanent files across several years, activity log with reminders. | – | Archive the version. |
| Quality management | ISQM-CH 1 with eight components, evidence from the audit data, inspection with report, annual evaluation; ISQM-CH 2 at engagement level; organisation with persons, licences, regulations, insurance and FAOA notifications. | – | Assess, define measures. |
A status overview on every engagement page shows where the audit stands; the engagement cockpit summarises key figures, materiality, capital situation, progress per audit area, findings and open items. The documentation is created alongside: whoever works through the steps has the complete audit file at the end.
Where the AI assists
pruvia uses AI where text has to be formulated, documents read or connections sought. Specifically:
- Knowledge of the business: suggestions on the market, environment and particularities of the audited company for planning.
- Document check: the audit procedure is prepared on the basis of the stored documents, including account reconciliation: does the balance match the document, and if not, what is missing?
- Root-cause search: if the trial balance does not balance, the AI looks for the likely reasons, such as signs, clearing accounts or missing accounts.
- Inquiry: word questions per audit area, match answers to questions.
- ICS (ordinary audit): derive key controls from a process description.
- Meeting note: transcription and draft with summary, statements, open points and tasks.
- Technical and usage questions: an assistant answers questions on the SER, the CO and on operating the software, in the context of the current topic; the full text of the SER is stored in the application and linked from the audit areas.
- Import of scans: only if a PDF has no text layer does the AI read the figures; otherwise pruvia works locally and deterministically.
Where the AI deliberately stays out
The journal entry analysis with its 14 test patterns, the leadsheets, key figures, materiality, capital situation and the report wording are fixed rules without AI. No AI result enters a working paper without an approval step. The AI can be switched off per company; pruvia then continues to work with all deterministic automations. And it gives no audit opinion: the assessment of risks, findings and report remains with the auditor.
Data protection of the AI functions
The AI runs exclusively via data centres in the EU (AWS Bedrock, Zurich endpoint, model inference EU-wide) and is fail-closed: if the European route is unavailable, there is no AI answer and no processing in the USA. Identifying details are pseudonymised for text-based functions; for documents this is not technically possible, so the company decides whether to use the document check. Nothing is used to train models. Application data is held in Switzerland, each company in its own database, with mandatory two-factor sign-in and automatic backups. The AI is included in the price.
Standards and types of audit
pruvia covers the limited statutory examination under the Swiss Standard on the Limited Statutory Examination (SER, Art. 729 et seq. CO), the ordinary audit under the Swiss Auditing Standards (SA-CH, Art. 728 et seq. CO) with the modules fraud (ISA-CH 240), ICS audit, bank confirmations (ISA-CH 505), sampling and comprehensive report to the board of directors (Art. 728b CO), and special audits under PH 10 “Special transactions”: incorporation, subsequent payment of capital, capital increase and reduction including capital band and conditional capital, capital deficit, interim financial statements and revaluation in case of loss of capital, conversion, liquidation; plus the SER special reports on interim dividends and Art. 725a para. 2 CO. Legal forms: company limited by shares, limited liability company, cooperative, association, foundation and others. Quality assurance follows ISQM-CH 1 in the audit firm and ISQM-CH 2 at engagement level.
Frequently asked questions
Does the AI replace the auditor?
No. The AI prepares and suggests; every adoption into a working paper is a deliberate step with initials. Audit opinion, risk assessment and report remain with the auditor, as the SER and SA-CH require.
How high is the degree of automation?
Everything that can be derived from the financial statements, journal and documents is created without manual work: leadsheets, key figures, materiality, capital situation, audit programme, journal entry analysis, report preselection, delivery and file. Manual work remains where judgement is needed. In practice, this means the documentation is created while you audit, and at the end there is no piecing together from Word, Excel and PDF.
Which data goes to the AI?
Only what the respective function needs, pseudonymised as far as technically possible, and exclusively to data centres in the EU with a Zurich endpoint. None of it is used for training. The function can be switched off per company.
Is the journal entry analysis AI-supported?
No, deliberately not. The 14 test patterns of journal entry testing are fixed rules so that the same data always yields the same findings and the derivation remains traceable in the file.
What happens if the AI is unavailable?
pruvia shows a short note, and the step can be completed manually. All deterministic automations continue to run independently of the AI.
Try pruvia
The system tour shows pruvia with sample data, without registration, including AI suggestions on an audit procedure. For questions about your audit firm, book an online demo.
Further Knowledge pages
Audit software Switzerland
What audit software for Swiss audit firms must be able to do – point by point with the current state in pruvia.
Journal entry analysisJournal entry testing: 14 test patterns
Which patterns pruvia looks for in the journal, what they find and how findings feed into the risk assessment.
Type of auditLimited statutory examination
Workflow under the SER, negative assurance, audit programme with leadsheets and auditor’s report – and what pruvia takes care of.
Type of auditOrdinary audit and ICS
Risk assessment, ICS existence check, tests of controls and comprehensive report to the board of directors – how pruvia maps them.
CollaborationDocument list and client portal
Request documents and questions with a deadline, control reminders and file incoming items as evidence in the audit area.
Quality managementISQM-CH 1 in the audit firm
Eight components, evidence from the audit files, monitoring, annual evaluation, as well as persons, regulations and insurance.
TermsAudit, Revision, examination: terms
What audit, “Revision”, statutory auditor, audit tool and audit software mean in Switzerland and how they relate.
Change of softwareTransition year and prior years
Mark years audited in the legacy system, file old audit files and start the first year in pruvia with previous-year figures.