swissTech LLC is a principal-led consulting practice building agent-driven analytics systems on Microsoft Fabric. We ship specification-first pipelines where AI agents read the data, write the narrative, and leave an audit trail executives can defend.
For two decades, BI assumed a human would stitch numbers into decisions. AI agents now do the stitching — faster, more consistently, and with a full audit trail. The dashboard doesn't disappear. It stops being the destination.
Reports surface the numbers. A human pivots, filters, and tries to reason a story into existence before the meeting starts.
Agents read the semantic model, reconcile with business rules, write the narrative, and cite every source. Humans review and approve.
Our differentiator isn't that we use AI — it's the discipline we apply to AI-native delivery. Four principles we hold to on every engagement.
Intent before code. Every deliverable begins as a specification precise enough that an AI agent can execute extended work sessions against it without drift. Humans review the spec; the agent runs the build.
Pipelines, not notebooks-left-behind. Every output regenerates from source with identical results. No "only Walter knows how to run this" artifacts ever leave our door.
Every agent action, every transformation, every generated narrative traces back to its source. Compliance teams follow the thread; executives defend the answer.
If a VP can't understand why the system said what it said, we haven't finished. Narrative layers translate technical decisions into plain language — by design, not afterthought.
We work where AI leverage is real, where enterprise platforms demand rigor, and where the alternative — a six-figure consulting pyramid — tends to disappoint.
lakehouse · pipelines · semantic_models
End-to-end builds: Lakehouse architecture, ingestion pipelines, Direct Lake semantic models, and the operational discipline to keep it running long after we're gone.
data_agents · storytelling · narrative_pipelines
Claude-powered data agents that answer business questions in natural language against your semantic models. Storytelling agents that generate executive narratives directly from Fabric data — cited, reproducible, reviewable.
specification_driven · agent_orchestration
When a process is repeatable, it should run itself. We build specification-driven workflows — batch analytics, regulated-domain automation — where the AI does the work and humans approve the outcome.
You will not be handed off to a junior resource. The person who scopes the engagement is the person who builds it — which is why a small studio delivers what larger firms quote in months.
A proposal from us is a short, specific document: the problem as we understand it, the approach, the deliverables, the price, the timeline. No decks. No vague "partnership" language.