Skip to content
    MICROSOFT COPILOT STUDIO · CONSULTING & DEVELOPMENT

    Copilot Studio consulting.
    Your agent. Your data. Your rules.

    We consult on, build and operate AI agents with Microsoft Copilot Studio — from the first use case to a production agent in Teams, on your website or in your CRM. Grounded in your content, with human escalation. From Zurich.

    Book an intro callSee our own agent
    Typically 2–4 weeks to productionGrounded in your dataConsulting & operations from Zurich
    WHAT IT LOOKS LIKE

    Configured in the open, not hidden in code

    Instructions, model choice (GPT or Claude models), knowledge sources: in Copilot Studio, every rule of the agent is there to read. We build it for you — your team can follow every adjustment afterwards.

    An agent's instructions, model selection and knowledge sources — source: Microsoft Learn documentation

    Getting started: describe the agent, have it built, refine — source: Microsoft Learn documentation

    WHERE AN AGENT PAYS OFF

    Four use cases we build most often

    Start with one clearly scoped case — not with “AI for everything”. In the intro call we pick the one with the fastest payoff.

    Internal knowledge assistant

    Policies, HR questions, IT support — the agent answers what circulates by email today, with sources from your approved content.

    Support chat with escalation

    Answers customer questions around the clock and hands over to a human as soon as it matters.

    Lead qualification on your website

    Qualifies enquiries in the chat and books an intro call directly — instead of a form and waiting.

    Customer service with Dynamics 365

    A service agent that creates cases directly in Dynamics 365 — no copy-paste between tools.

    MORE THAN A CHATBOT

    What separates an agent from a chat window

    A public chatbot knows the internet. Your agent knows your policies, your prices, your processes — and its limits.

    Grounded in your data

    SharePoint, Dataverse, websites, APIs — the agent answers from approved content and can show its sources.

    Response policies

    You define what the agent may say and what it may not. Not a vibe — a rule.

    Authentication built in

    Microsoft Entra ID or external identity providers — the agent knows who it is talking to.

    One agent, many channels

    Teams, web widget, mobile app or directly in Dynamics — built once, deployed where your people are.

    Human escalation

    Critical decisions belong with people. The agent hands over — with the full conversation context.

    Evaluated before go-live

    Generative answers get tested and rated before anyone sees them — and monitored after launch.

    WE USE ONE OURSELVES

    The Theia Support Agent answers our customers' requests

    In our Managed Service, an AI agent handles first-level requests and escalates to a human when needed. We don't sell agents we wouldn't run ourselves — have a look at how it works.

    See the Support Agent
    HOW IT WORKS

    Three steps to a production agent

    Typically 2 to 4 weeks from kickoff to production — the exact plan is fixed in writing in the intro call.

    1

    Intro call (30 min, free)

    We pick the use case with the fastest payoff, clarify data sources and channels — and you receive a written fixed-price quote for the implementation.

    2

    Build, ground, evaluate

    We build the agent, connect your content, define response policies — and evaluate the answers before anyone sees them.

    3

    Go-live & operations

    Deployment in Teams, on the web or mobile. On request we keep operating the agent: monitoring, policy adjustments, new data sources.

    Book an intro call
    Beyond the agent

    The AI workflow for SMEs

    One workflow, clearly scoped, with evaluation and human sign-off — typically live in 4–8 weeks.

    For the technical readers

    Copilot Studio in detail

    Features, use cases and official Microsoft resources on our technology page.

    FAQ

    Common questions about Copilot Studio agents

    That depends on the use case — data sources, channels and integrations vary widely. After the free intro call you receive a written quote with a fixed price for the implementation. Microsoft's licence costs for Copilot Studio come on top; we walk you through them transparently before you decide.

    Typically 2 to 4 weeks from kickoff to production — depending on data sources, integrations and approvals. We put the exact plan in writing in the intro call.

    Microsoft's low-code tool for building your own AI agents. The agent is connected to your data — SharePoint, Dataverse, websites, APIs —, gets response policies and then runs in Teams, on the web or on mobile. You don't need your own LLM-ops team for it.

    The risk can be reduced, but not argued away: grounding anchors answers in approved content and can show sources, evaluations check quality before go-live, and for critical cases the agent escalates to a human. That is exactly how we build every agent.

    Not much: a Microsoft tenant and the content the agent should rely on — SharePoint documents or your website, for example. Sign-in and permissions run via Microsoft Entra ID or external identity providers. Whether your existing licences are sufficient is something we clarify in the intro call.

    Yes, on request: monitoring, policy adjustments, new data sources and further development. The agent runs in your tenant and belongs to you — there is no lock-in either way.
    30-MINUTE INTRO CALL

    Your agent, sketched out — before you commit.

    In 30 minutes you know which use case pays off first, which data the agent needs — and what it costs. In writing, no obligation.

    Book an intro call

    We use cookies and external services (e.g. Google Maps) to provide you with the best experience on our website. For more information, see our Privacy Policy.