AI agents and automations that actually work
We build conversational agents and automations that operate on your real data and integrate with CRM, ERP and back-office systems. A working prototype lands in weeks four to six: you see it work before investing the rest.
What we build
- Conversational agents (RAG): assistants answering on your company knowledge base, in multiple languages, on your site and sales channels. For customer care, internal helpdesk and sales support.
- Document automation: quotes, contracts, reports and communications generated and checked automatically from your data and templates.
- Virtual sales agents: they qualify inbound leads, follow up, prepare appointments and stay active 24/7.
- Integrations: CRM, ERP, back-office, e-mail and messaging channels. Data pipelines and observability from day zero.
How the project runs
We start from a single process with enough volume to justify the investment. In the first weeks we analyse the process and data, then build the prototype and put it in the hands of the people who will use it daily: we iterate until the frontline approves. After validation we move to release with integration into existing systems and key-metric monitoring. On average, from kick-off to release it takes 6 to 14 weeks.
If you wish, the project closes with a technology transfer: full documentation, training for your team and support until the system is manageable in-house. Our stated goal is to become superfluous.
Security and governance
We work on-premise, in private cloud or with GDPR-compliant European providers. No company data is used to train third-party models without explicit consent. Every agent ships with logs, answer-quality checks and documented action limits, so AI Act compliance is built in from the design phase.
A concrete example
Our best-known project is Langhe My Food, the AI system guiding tourists across the Langhe and selling local products with an AI guide for pairings and cooking: category winner and third overall at the Assoprom Awards 2026. The same logic, adapted to your process, is what we apply in agentic projects.
Q.1What do we need to start?
A process with repetitive volume, access to the data or documents involved and an internal owner who knows the process. We bring everything else: analysis, prototype, release.
Q.2Where does our data go?
We work on-premise, in private cloud or with GDPR-compliant European providers. No company data is used to train third-party models without explicit consent.
Q.3How long is a project?
Typically 6 to 14 weeks from kick-off to release. The working prototype on real data lands in weeks four to six: you see it work before investing the rest.
Q.4Who runs the system after release?
You, if you want. Technology transfer is part of the method: we train your team with documentation and support until the system is manageable in-house.
Q.5Is the initial pre-analysis really free?
Yes: the pre-analysis we run together at the start, to assess feasibility and scope, never has a cost. The prototype described above, and any demo, are later paid phases, priced according to the project's complexity and defined in the proposal.
Got a process to automate?
Bring us your most tedious process: if automation does not pay off, we will tell you before we start.