AI that drafts, writes and prepares — inside the tools you already use.
Writing assistants, document and proposal generation, copilots wired into your business tools. The right model for each task, with guardrails your data deserves.
Generative AI produces text, documents and answers on demand: drafting emails and proposals, generating reports from your data, assisting teams inside their business tools. Deployed well in a company, it means choosing the right model per task — Claude, ChatGPT, Gemini, Mistral, Llama — wiring it into real workflows, and setting guardrails for sensitive data.
For Swiss SMEs and growing companies whose teams already paste things into ChatGPT — without rules, without connection to internal data, and without knowing what leaves the building. And for those who have not started, but see competitors drafting proposals in minutes.
Generative AI has already entered your company — as a browser tab someone pastes client emails into. The draft is decent, but the AI knows nothing of your offers or your tone, and nobody controls what just left the building. The gap between that and a real deployment is not the model; it is everything around it.
We build assistants wired into your actual tools — CRM, ERP, email — grounded in your documents and your voice, with the right model per task among Claude, ChatGPT, Gemini, Mistral and Llama. Guardrails come first: what the AI may see, what a human must validate, what routes to local AI because it must never leave. Then we train your teams, because that is where the multiplier is.
From playing with ChatGPT to working with generative AI
Most companies have already met generative AI — as a browser tab. Someone pastes a client email, gets a decent draft, and moves on. Useful, but shallow: the AI knows nothing of your offers, your tone, your history with that client, and nobody controls what was just pasted into an external service.
The productive version looks different. The assistant is wired into your tools, reads the right context from your own documents, drafts in your voice, and operates under written rules. The difference between the two is not the model — it is the integration around it.
What we build with generative AI
We focus on use cases with measurable time savings, not demos.
- Proposal and quote generation from your catalogue, past offers and client context
- Email and correspondence drafting in your tone, multilingual — French, German, English
- Report and summary generation from meetings, documents or your own data
- Copilots inside the CRM or ERP: the assistant appears where the work happens
- Document assistants grounded in your internal knowledge — connected to enterprise RAG when depth matters
Choosing models: never a one-vendor bet
No single model is best at everything. Claude excels at long, careful writing; ChatGPT is a strong all-rounder; Gemini handles large contexts well; Mistral and Llama run locally when data must not leave. Pricing, speed and quality shift every quarter.
So we architect for choice: your assistants sit on a layer where the model behind each task can be swapped as the market moves. You are never locked into one vendor's pricing or one vendor's outages.
Guardrails and sensitive data
Generative AI without rules is a liability: confidential data pasted into external services, invented figures in client documents, tone drifting off-brand. We set the guardrails first — what the AI may access, what must be verified by a human, what may never leave your infrastructure.
For genuinely sensitive data — client files, medical or financial records — we route those tasks to local AI running on your servers or in a Swiss cloud, so confidentiality holds by construction, in line with nLPD. And we train your teams, because a well-briefed team gets far more out of the same tools.
- •Writing assistants grounded in your documents and tone
- •Document generation: proposals, reports, contracts, summaries
- •Copilots inside your business tools — CRM, ERP, email, GED
- •Model selection per task: Claude, ChatGPT, Gemini, Mistral, Llama
- •Guardrails: what the AI may see, say and never touch
- •Local AI routing for sensitive data, nLPD-compliant
- •Working assistants integrated into your actual tools, not a separate chat page
- •A model architecture that lets you switch providers as the market moves
- •Written guardrails: access rules, human validation points, data boundaries
- •Sensitive-data routing to local AI where confidentiality requires it
- •Your teams trained on the assistants, with usage rules that stick
The value is in the integration, not the chatbot tab.
Right model per task: Claude, ChatGPT, Gemini, Mistral or Llama — never one vendor by default.
Guardrails come first: access, validation, data boundaries.
Sensitive data routes to local AI — nLPD by construction.
Trained teams extract several times more value from the same tools.
Smart CRM
AI augmented CRM, built for your business, wired into your tools.
AI agent / assistant
AI agents that act inside your tools, not just chatbots that answer questions.
Enterprise RAG
RAG on your knowledge base so AI answers with your context, not in general.