AI & Agents — the path

03Automate

Where off-the-shelf tools stop, agents start.

Copilot helps your team. Agents work for your company: they take calls, answer customer questions from real company knowledge, and handle back-office tasks — with their own identity, auditable actions, and human approvals. Three systems running today:

In production. Today.

Three systems. No slides.

In customer use

The AI receptionist callers can't tell is AI.

It takes real calls, switches language with the caller, and answers from the company's own knowledge base. It reads employees' Teams presence, routes calls to whoever's available, takes messages — and, on request, emails the caller a summary of the call.

  • No lost calls — routed by Teams presence, or a message taken
  • Call summary emailed to the caller
  • Answers from real company knowledge — the same knowledge base as the chatbot
Live call · +41 ·· ··· ·· ··

Caller

«Hello — I need to move my Thursday appointment.»

Language detected: Swiss German

Agent

«Sure — I see your appointment at 10am. Friday at 2pm is open — does that work?»

Teams presence checked: case handler available — transfer offered

Appointment moved — summary emailed to caller00:47

What she sounds like

Listen to the AI receptionist (AI-generated example)

0:00 / 0:00
In customer use

The chatbot that knows the company. Not the internet.

Most chatbots guess. This one answers from the company's central knowledge base — Azure AI Foundry with Foundry IQ, hosted in Swiss Azure regions. The same knowledge base the receptionist draws on: maintained once, consistent everywhere.

  • Correct answers at 11pm instead of a queue the next morning
  • Noticeably fewer routine tickets for the team
  • Data stays in Switzerland — nDSG-compliant
Web chat · 11:02pm

Do you also deliver to Austria?

Yes — deliveries to Austria take 2–3 business days, free shipping on orders over CHF 200.

Source: Shipping policy · p. 2

And special terms for resellers?

Outside its knowledge — handed off to the sales team ✓

Internal, every day

Cleo — the executive-assistant agent we run our own company on.

Inbox triage, timesheets, pipeline monitoring, briefings, content drafts: Cleo handles our own company's back office — among other things. Her own identity, every action traceable, approvals stay with humans. We don't sell anything we don't run ourselves.

  • Back-office hours reclaimed for real work
  • Every agent action traceable to an actor
  • Approvals stay with humans — always
Agent operations · today
Inbox triaged — 14 emails07:45
Timesheet prepared — 4 entries16:45
2 pipeline follow-ups draftedApproval: human17:00
Audit log written17:06

End of day: 0 open emails · every action traceable to an actor

What she sounds like

Listen to Cleo's morning briefing (example, 30 sec.)

0:00 / 0:00

From the field

Receptionist and chatbot for a Swiss trade association

  • Swiss trade association
  • Phone and web enquiries, around the clock
  • Azure AI Foundry + Foundry IQ
  • Live in production, 2026

Starting point

Recurring member enquiries by phone and web tied up capacity — outside office hours, much of it went unanswered.

Approach

A chatbot with real association knowledge (Foundry IQ) and a multilingual AI receptionist (ElevenLabs voice) — both in Swiss Azure regions, with a defined handover to humans.

Result

Enquiries get answered around the clock; whatever the AI isn't sure of lands with the team, cleanly documented. The guardrails were half the work — deliberately.

Anonymised. Reference available on request during the intro call.

The foundation of every agent

  • Azure AI Foundry + Foundry IQ: answers from your knowledge, not the internet
  • Swiss data residency — nDSG-compliant, your tenant, your sovereignty
  • Entra Agent ID: every agent with its own, controlled identity and audit trail
  • Human approvals for anything that sends, commits, or pays
  • The guardrails are the product: defined, configured, documented

The path to your own agent

  1. 01

    AI Assessment — CHF 2,500

    Data landscape, processes, tenant readiness. Use cases prioritised by impact — and an honest word on what isn't worth doing.

  2. 02

    Pilot with guardrails

    One use case, close to production, with security gates. Measured against a baseline, not enthusiasm.

  3. 03

    Operations

    Rollout, monitoring, audit, ongoing development — as a managed service.

Where does your company stand today?

Next step04Operate — Worxspace Managed