Frequently asked questions
Everything people ask before an audit.
General
What can AI concretely do for a Swiss SME?
AI automates repetitive tasks that weigh on your team: sorting and replying to recurring emails, generating standardised documents, extracting data from PDFs, classifying support tickets, drafting quotes. For an SME of 10 to 50 people, that typically means 5 to 15 hours per week freed up. AI doesn't replace your domain expertise: it removes low-value administrative work so your team can focus on the rest.
Do I need a data science team to get started?
No. Modern tools (GPT, Claude, n8n, Make, Dify) let us ship useful projects without training a model or hiring a PhD. Our role is precisely to avoid unnecessary complexity: we plug existing models into your data and your tools via standard integrations. You don't need a data lake, MLOps, or a dedicated team. An executive sponsor and one technical person who knows your internal tools is enough.
What about data confidentiality and the Swiss nFADP?
The new Swiss data protection law (nFADP, in force since September 2023) strictly governs the use of personal data, including its transfer outside Switzerland. By default we work with providers hosted in Switzerland or the EU, and we exclude sensitive data from prompts sent to public LLMs. For cases that require it, we deploy self-hosted models on your infrastructure or in a Swiss datacenter. Every project starts with an inventory of the data processed and an explicit decision on where it flows.
How long before the first concrete result?
A first workflow running in production generally takes 3 to 6 weeks after kick-off. The initial audit (2 weeks) identifies a short use case, the build phase (2 to 4 weeks) delivers a first version your team can actually use. We prefer shipping a small thing that works than a big project that never lands. Further iterations come later, based on real feedback.
Project
How does an AI audit work?
The audit starts with a 30-minute call to understand your context, your tools and your pain points. If the lead is relevant, we run an on-site (or video) workshop of 2 to 3 hours with the people who do the work day-to-day. You leave with 3 prioritised recommendations, an impact estimate (hours saved, cost, timeline), and a roadmap if you want to go further. You keep a usable document, even if you don't continue with us.
What's the indicative price range?
A one-off project sits between CHF 1,500 and CHF 15,000 depending on complexity. Simple workflow automation or an internal chatbot on documentation: CHF 1,500 to 4,000. AI integration into an existing business tool with connections to several systems: CHF 5,000 to 10,000. A full RAG project with a document base and multiple use cases: CHF 10,000 to 15,000. The initial audit (30 min + workshop) is free and without commitment. Ranges are given after scoping, not before.
What is your 3-phase method?
Phase 1 — Audit: we map your processes, identify where AI has real leverage, and prioritise by impact/cost ratio. Phase 2 — Build: we deliver a functional V1 in 2 to 4 weeks, tested with your teams, documented. Phase 3 — Run: we deploy, train the team, monitor. Each phase has a clear deliverable and a go/no-go. If the audit shows AI isn't the right answer to your problem, we'll tell you; we don't sell projects that make no sense.
Is maintenance included after the project?
Yes, 3 months of maintenance are included in every one-off project: bug fixes, minor adjustments, Q&A. Beyond that, you choose: take over internally (we train you and deliver documented code), subscribe to a monthly retainer, or call us on demand. We don't manufacture artificial dependency: the code is yours, the prompts are yours, you can leave whenever you want.
Technical
Which tools do you use?
LLMs: GPT (OpenAI), Claude (Anthropic), Mistral for sovereign cases, Llama for self-hosting. Orchestration: n8n for internal workflows, Make for quick integrations, Dify for RAG chatbots, custom code (Python/TypeScript) when no-code hits its limits. Infrastructure: Supabase, Postgres, vector stores like Qdrant or pgvector. We pick the tool for the need, not the other way around — and we avoid proprietary stacks that lock you in.
Can data stay in Switzerland?
Yes, it's a common requirement for our clients. LLM providers now offer European regions (Azure OpenAI Swiss North, AWS Bedrock Zurich, Mistral in France). For self-hosting, we deploy on Infomaniak, Exoscale or your on-premise infrastructure. Orchestration (n8n, database, vector store) runs in the region of your choice. We precisely document where each piece of data flows in a diagram you can show your DPO or your clients.
Does AI integrate with our existing stack?
Yes, it's 80% of the real work. We plug AI into your current tools via their APIs: Microsoft 365, Google Workspace, Salesforce, HubSpot, Bexio, Abacus, SharePoint, your business databases. If a tool has no API, we use authorised internal scraping or file import. The end user doesn't see yet another new tool: they see a button in their CRM, a script in their Excel, or a message in Teams. Adoption depends on that invisibility.
Will my employees keep their jobs?
In the projects we've delivered, no position has been cut. AI takes on tasks nobody was doing well for lack of time: follow-ups, qualification, data updates, reporting. Teams get back 5 to 15 hours per week that they reinvest in commercial, relational and creative work — the work that actually generates value. If your intent is to replace headcount, we're not the right partner; we work with SMEs that want to do more, not less.
Another question?
30 minutes, one call, zero commitment. We'll tell you if AI is the right answer to your problem — or not.