What the CFO is accountable for
Control, accuracy, audit defensibility, timely numbers for decisions and a finance function that scales more slowly than headcount and revenue. What the CFO distrusts is black-box automation, numbers that cannot be explained, migration risk and vendors who talk about AI without showing the ledger. Aderis is built around the first list and answers the second with evidence rather than reassurance.
The control model: rules before execution, exceptions after
In an agentic finance system, control does not mean approving every step by hand. It begins with defining what an agent may execute on its own and where it has to stop. Two business partners with the same tax number, an amount above the approval limit or a missing account assignment are not cases for assumptions; the agent halts, documents the deviation and escalates it with context to the responsible person. The CFO sets permissions, approval limits, segregation of duties and escalation paths; agents execute within that frame. Interpretative AI and deterministic posting logic remain separate, so the entry that results is reproducible. Every action remains traceable to its source, the rule applied and the approval given, and a confidence score on every automated posting shows what an agent handled on its own.
Current numbers, consolidated
Continuous Close keeps the general ledger, reconciliations, journals, accruals, assets and VAT current throughout the period, so the question "are the numbers complete?" gives way to "what do they mean?". Multi-entity, multi-ledger and multi-currency run on one core: statutory, tax, management and group ledgers in parallel from a single source of postings, with consolidation on the same base rather than reassembled in a spreadsheet. Ada, the conversational finance agent, answers questions such as "consolidate today's figures for all entities in CHF" or "what is holding up the close?" from the live financial context and can carry the next step forward within the CFO's permissions.
Audit and tax
Audit trails, approvals and segregation of duties are on by default and part of every workflow. Every figure stays linked to its supporting document, decision and approval throughout the retention period, so a revenue reconciliation for a year long past can be produced from the record rather than reconstructed. Known sources of VAT error, such as missing VAT on intragroup transactions, advance payments or an incorrect place of supply, can be turned into validation rules that agents apply to every entry during the period, so the annual VAT reconciliation confirms work already done. The platform is designed to meet GeBüV and GoBD requirements, data is hosted in Switzerland and the EU, and ISO 27001 certification is in progress.
Migration risk
Aderis treats migration as a service rather than a project. The Aderis team connects banks and source systems, maps the chart of accounts and migrates master data, opening balances and history, with every figure traceable back to source; in documented implementations this covered seven to eight years of history. From kickoff to live finance is measured in days, and the transfers are included. The company goes live with reconciled numbers rather than balances taken on trust.
What to verify before deciding
Eight questions separate execution depth from AI marketing, and Aderis expects to answer them in a demo. Which process steps does the system execute itself, and which does it merely prepare? What data does the AI work on, and how current is it? How does the system detect and handle uncertainty, exceptions and conflicting information? Can every figure be traced back to the original document, the transaction and the decision behind it? Where are roles, permissions, segregation of duties and approvals technically embedded? How are multiple entities, ledgers, charts of accounts and currencies managed? Which interfaces are standardised, and where do manual file exports remain? Which operational metrics will measure the benefit after 90 and 180 days?
Frequently asked questions
Is this a black box?
No. Every entry shows whether an agent or a person posted it, which rule applied, what the confidence was and who approved it. The chain from report to document is intact and visible on the platform.
How disruptive is the migration?
The Aderis team runs it, including the data transfers, and it is measured in days rather than quarters. Opening positions are reconciled before the platform takes over day-to-day accounting.
Where is our data, and who can access it?
In Switzerland and the EU, encrypted in transit and at rest. Access is permission-based with full audit trails; agents are subject to the same permissions as people.
What changes for my team?
The routine moves to agents; the team sets rules, handles exceptions, approves where judgement matters and works with current numbers. Capacity moves from data preparation to interpretation.
