Thinks in-house.
The language layer runs on the Apple chip – no cloud, no subscription. What you ask and what the assistant answers stays in-house.
AI that supports your work.
Classify documents, spot anomalies and ask questions about your numbers. You choose which AI features to enable and how to run them. Suggestions remain traceable and approvals stay with your team.
Questions about your business
Ask questions. Put figures in context.
Which invoices are overdue? Where do you stand with a customer? The assistant uses your system’s business data and permissions. Sources and evidence cards help you follow the answer.
It only reads. You confirm.
"Send reminders to the overdue debtors" prepares the dunning run – nothing is sent until you confirm in the familiar dialog. The same goes for the period close. The assistant never books, pays or sends anything itself.
Explains the payroll.
For a question about a pay change, the assistant compares the two most recent payslips by component and highlights the largest movement. Payroll staff gain a starting point for their review.
Knows the receipt inbox.
"What is in the receipt inbox?" – how many receipts are waiting, how many are ready for conversion, how many need your review. Conversion happens in the OCR inbox, after your confirmation.
Briefs trustees.
"Which mandate needs attention today?" – the morning triage across all mandates, most urgent first, with the francs at risk.
Your language, one price.
You ask in German, French or Italian, the answer comes in the language of your interface – and the assistant is part of ewws: the edition prices remain unchanged.
Answers the owner's question.
"Is there enough money?" – the daily briefing shows the liquidity outlook as a traffic light with the amount in francs and flags what needs your review before the next bank reconciliation. Deterministic, straight from your figures, no model run needed.
Collects your decisions.
Pending payment releases and deletion requests appear as a card – with amount, requester and a jump to where you decide under the four-eyes principle. The assistant prepares; you confirm in the module.
Comes along on the iPhone.
ewws AI also runs in the iPhone app from the App Store – on-device, with the same evidence cards and permissions as at your desk. Your books stay in-house, even on the road.
The language node
One Mac mini brings the AI to the whole company – Windows included.
In a mixed Mac and Windows environment, a Mac with Apple Silicon can handle local language processing. Windows workstations access the configured language node through the ewws server.
One device thinks for everyone.
The Apple Foundation Models run on the Neural Engine of one Mac on the company network – a Mac mini is enough. A switch in the ewws settings turns it into the language node; no additional licences, no AI subscriptions.
Windows asks, the node thinks.
The Windows workstation asks in the assistant. The ewws server relays the question to your tenant's language node and carries the answer back: question and answer stay within your installation, the thinking happens on the node.
Visible, not just claimed.
The cockpit shows the language node as a presence signal: green when it is ready, amber when it has not reported for a while. And every answer carries its source – which tool, which figures.
The local language feature requires a suitable Mac with Apple Intelligence available and a configured language node on your company network. The node is linked to your tenant; Windows shows when it is unavailable.
Data sovereignty first
Your data belongs to you. Even when AI helps.
AI document processing starts disabled. You deliberately choose activation and the operating mode: locally on a compatible Mac, through a cloud provider with your API key, or as an individually planned integration.
Start switched off. Choose the data path.
Document AI starts switched off. The mode selected on a Mac applies to all its tenants. Reminder recommendations, forecasts and payroll checks use their own deterministic rules and do not need a language model.
Integrate your own AI (Bring Your Own AI)
Other providers or your own inference server are not a preconfigured standard option. Such an integration requires a separately planned development project. Providers, data flows, contracts and operational responsibility must be agreed beforehand.
Professional confidentiality
Choose how your data is processed.
With local document processing, content stays on the Mac. If you enable a cloud provider, the content needed for processing is sent to that provider. Clarify the mode and requirements for your business before activation.
Fiduciary offices
With local document processing, client documents are classified on the Mac. Your team can review suggestions and configure processing around the agreed data flows.
Law firms
Fee invoices and expenses can be processed locally on the Mac. Your firm can use AI assistance without sending those document contents to a cloud provider for classification.
Medical practices
Supplier invoices and expenses can be classified locally on the Mac. Practice staff review the suggestions and decide what to carry into accounting.
Associations and foundations
Small teams can also classify documents locally on the Mac. This processing does not require an additional contract with a cloud AI provider.
Self-hosted means: the ewws server stands on your hardware or in your public cloud in your own account. On device additionally means: the AI model runs on the Neural Engine of the Apple Silicon Mac, not via a server request to an LLM provider. BYOK (Bring Your Own Key) means: whoever activates Anthropic Claude or OpenAI signs a contract directly with the provider and stores their own API key in ewws. We do not route through a collective account and do not see the contract content.
Operator overview
What the AI wants today, you see immediately.
Open recommendations, uncertain cases and the OCR adoption rate appear in one overview. Your team can see which documents and suggestions need attention and choose the next task.
Important today at a glance
- Top AI recommendations with severity (critical, review, info), sorted by impact.
- An uncertain queue with the cases from all modules that have low confidence.
- The OCR adoption rate of the last 30 days, with top suppliers and correction share.
Cross-module recommendations
Four recommendation sources, one place. Every recommendation with evidence.
The AI connects OCR, dunning, banking and reordering into a shared stream of suggestions. Not "the AI thinks", but a concrete statement with a sample size and a before-after comparison.
What the four recommendation sources deliver
- The OCR correction trend per supplier, when the operator increasingly corrects posting suggestions.
- Dunning clusters across several customers with the same risk profile.
- A reorder correction trend per article as well as a banking review trend per IBAN, each with a delta over 30 days.
AI question
Ask your books, even entirely without a language model.
You type the question in German, French or Italian: what is open, who has owed the longest, how liquidity stands, what the VAT quarter looks like. ewws computes the answer from your own figures, names the basis it came from and suggests the next step. Without a language model, without cloud, and even when no Mac is on the network.
Eighteen questions ewws answers without a model
- Cash and risk: open receivables/payables, critical dunning, cash-flow horizon, solvency risk, largest overdue customer.
- OCR and AP: supplier with the most overrides, documents posted automatically today, confidence trend.
- Operational: stock warnings, top customer revenue, payroll run, VAT quarter, open expenses and time entries. You ask in German, French or Italian; the answer appears in the language of your interface.
Payroll anomaly check
Detect anomalies before the ESTV submission, not after.
Five plausibility detectors check every payroll run before submission: gross jump, deduction rate, employer-cost rate, net plausibility and new employees. Every anomaly comes with a comparison value from the last periods.
What the detectors detect
- A gross jump per employee with a comparison to the median of the last periods.
- Conspicuous deductions or an employer cost rate in individual payslips.
- New employees without a historical reference and net plausibility against gross.
From the application
AI that justifies its recommendation
No black box: every recommendation shows its evidence - critical items, scores, open amounts and overdue status. And the dunning score rates every open item individually, with a justification and a per-line breakdown.
Operating modes
How should your documents be processed?
Choose the data flow that fits your business. Local and cloud processing have different requirements. Document assistance remains disabled until you activate it.
| Mode | Processing | What you need |
|---|---|---|
| Disabled | No AI classification of your documents. | The initial state. Normal ERP workflows remain available. |
| Locally on Mac | Apple Foundation Models processes document content on the device. | A compatible Mac with Apple Intelligence available and the feature enabled. |
| Cloud with your own key | Required content is sent to your chosen provider: Anthropic or OpenAI. | Your own API key, an appropriate agreement with the provider and explicit activation. |
| Custom integration | Data flows and processing are defined for your integration. | A separate development project, not a preconfigured standard option. Data flows and operational responsibility are agreed beforehand. |
Local document processing and the language node for Windows are separate features. The language node requirements are described in its section above.
What lies ahead
Honest about what the AI cannot do yet and when it is expected.
No AI hype. No promise that shines in the next demo and disappoints in everyday use. These two areas are next in line, with their own iterations.
Industry-specific anomaly profiles
The payroll detectors are generic. Hospitality and construction need different thresholds, for example because of 13th-month-salary advances, vacation payouts, hourly-wage models. Industry profiles for hospitality and construction follow with their specific thresholds and wage models.
An optional LLM layer for summaries
For customers who explicitly want LLM summaries, an optional layer is offered. Voluntary activation, with a clear record of which data goes where and to which provider. By default it is and stays off.
How you get started
Three steps from off to an AI you can follow.
AI document processing is initially switched off. The selected mode applies to all tenants on that Mac. You deliberately choose where document text is processed.
Step 1
Choose the route
Choose the mode in the Mac app: off, local processing on a suitable Mac, or a supported cloud provider with your own API key. The operating-mode overview explains data flows and requirements.
Step 2
Ask, and check the suggestions
The assistant answers questions about your books and puts forward suggestions: dunning clusters, posting proposals from the document inbox, replenishment. Every suggestion shows what it rests on, with items, amounts and deadlines beside it.
Step 3
Decide, and only then does it happen
The AI reads and calculates; posting and sending happen only after your confirmation. What you confirm runs through the same path and the same rights as work done by hand, and it stands in the same audit trail.
AI consulting and data sovereignty
Use AI without giving up control of your data.
We have worked with neural networks, LLMs and local AI architectures for many years. That experience goes beyond ewws. We help Swiss companies choose viable use cases, define data flows deliberately and introduce AI while keeping responsibility and decisions with people.
From a problem to a viable use case
We assess which task genuinely benefits from AI, which data is actually required and where a deterministic solution is more reliable. The result is a sound basis for a decision, not an AI project for its own sake.
The right architecture for sensitive data
On-device, self-hosted or deliberately using a cloud provider: we clarify models, data flows, interfaces, contracts and responsibilities before sensitive content is processed.
From decision to implementation
We support architecture, prototyping, integration and business acceptance. People retain approval and responsibility; the data path and result remain traceable.
Taking stock
Try AI on a real task.
Start with a question about a document or report. Check the answer and its sources, then decide which support fits your business.