More Information

Exosphere Agentic: governed agentic loops for BSS

Pick a loop. Improve an outcome.

Exosphere Agentic gives AI agents a shared telecom ontology, your own operating context and a governed path to act, over the BSS and OSS you already run. Exosphere Agentic runs on Azure and integrates with the Microsoft AI platform, deployed into the customer's own tenant — so context, data and governance stay where they belong.

Updated 11 October 2026

From system of record to system of outcome

Pick a loop. Improve an outcome.

Your BSS records what happened. Exosphere Agentic helps you act on it. It gives AI agents a shared telecom ontology to reason with, grounds that reasoning in your own products, contracts, policies and processes, and gives every action a governed path: eligibility, authority, approval and an audit trail.

A shared telecom ontology, your own context, and governed action, exposed over MCP.

Overlay, not rip-and-replace. Loops run over the BSS and OSS you already have, whoever built it. Your systems of record stay authoritative.
Agents run the loops. People govern. Governance is not a wrapper on top; it is the layer every call passes through.
Standards as the contract. Governed actions reach your systems through TM Forum Open APIs, so the approach stays backend-neutral.

Four layers, one governed path

flowchart TD
    subgraph SF["SURFACE"]
        C1["Microsoft 365 Copilot<br/>and Microsoft Teams"]
        C2["Web, care and<br/>partner channels"]
    end
    subgraph AG["MICROSOFT AI PLATFORM"]
        A1["Copilot Studio"]
        A2["Microsoft Foundry"]
    end
    subgraph EXA["EXOSPHERE AGENTIC"]
        E1["Telecom ontology<br/>shared meaning"]
        E2["Operating context<br/>your products and policies"]
        E3["Exosphere judgment layer<br/>policy · orchestration · audit"]
        E4["MCP and tool gateway<br/>TM Forum Open APIs"]
    end
    subgraph SOR["SYSTEMS OF RECORD · UNCHANGED"]
        S1["Billing and charging"]
        S2["CRM and ordering"]
        S3["Service and<br/>resource inventory"]
    end
    SF -->|"intent"| AG
    AG -->|"intent"| EXA
    EXA -->|"authorised calls only"| SOR
    SOR -.->|"evidence and outcomes"| EXA
    EXA -.->|"outcome"| AG
    AG -.->|"outcome"| SF
    classDef ms stroke:#4A5F7A,stroke-width:2px
    classDef exos stroke:#0077B6,stroke-width:2px
    classDef sor stroke:#7A8DA5,stroke-width:1px
    class C1,A1,A2 ms
    class E1,E2,E3,E4 exos
    class S1,S2,S3 sor

Intent flows down, evidence and outcomes flow back, and only authorised calls reach your systems of record.

What each layer does

Layer	What it does	Provided by
Surface	Where people ask, review and approve: Microsoft 365 Copilot, Microsoft Teams, web and care channels	Microsoft and your own channels
Agent platform	Where Sherpa agents are built and run, inside your tenant	Copilot Studio and Microsoft Foundry
Exosphere Agentic	Telecom ontology, operating context, loop orchestration, governance and authority, MCP and tool gateway	Exos Systems
Systems of record	Billing, CRM, ordering, product and service inventory, unchanged	Your existing stack

Inside Exosphere Agentic, Exosphere is the judgment layer: a shared entity model with one identity across APIs, intent resolved to capability, orchestration with sagas and idempotency, and policy with lineage from intent to record.

Microsoft provides the agent platform. Exosphere Agentic provides telecom meaning and governed execution.

Microsoft product names are used descriptively.

Every loop runs the same governed pattern

A loop is a closed cycle from signal to outcome, not a chatbot in front of a form.

Detect a signal from billing, ordering, usage or the network.
Understand it in telecom and customer context.
Decide under policy. This is where a person sits whenever policy says so.
Act once, through a governed tool.
Confirm the real outcome and record it.

Three rails run under every loop: Policy (what is permitted), Authority (who may act) and Evidence (why this action was selected). Every answer follows the Sherpa envelope: Facts, Assessment, Recommendation, Action, Audit. Missing data is stated, never filled in.

Start from the outcome

flowchart TD
    Q{"Which outcome first?"}
    Q --> PR["Protect revenue"]
    Q --> RC["Reduce cost to serve"]
    Q --> AR["Accelerate revenue"]
    PR --> PR1["Credit risk · Fraud exposure<br/>Revenue leakage"]
    RC --> RC1["Bill explanations<br/>Bill-shock prevention<br/>Enterprise invoice analysis"]
    AR --> AR1["Upgrade recommendations<br/>Enterprise quotes<br/>Contract renewals"]
    PR1 --> PRM["Measure: cash,<br/>leakage, exposure"]
    RC1 --> RCM["Measure: effort,<br/>resolution, backlog"]
    AR1 --> ARM["Measure: conversion,<br/>cycle time, time to bill"]
    PRM --> PATH["Pick · Contextualise · Connect · Prove"]
    RCM --> PATH
    ARM --> PATH
    classDef exos stroke:#0077B6,stroke-width:2px
    class PATH exos

Example loops only. The right first loop is the one tied to the outcome you most need to improve.

Start with one loop. There are many more.

Every loop follows the same governed pattern, which is what makes the first one worth starting and the next one cheap to add. The right loop to start with is the one attached to the outcome you most need to improve.

Step	What happens
Pick	Choose one material outcome and the measure that will prove it.
Contextualise	Extend the telecom ontology with your products, contracts, policies and terminology.
Connect	Connect only the systems that loop needs, through TM Forum Open APIs.
Prove	Measure before expanding, then reuse the foundation for the next loop.

The example loops shown above are a representative sample of the Exosphere Agentic catalogue, not the full set. New loops are engineered continuously as customers identify the outcomes that matter most to them. If the loop you need isn't listed, it can be built.

Explore Exosphere Agentic

The telecom ontology for BSS and OSS: how SID's shared nouns become meaning your agents can act on.
From cell alarm to credit note: where the network meets the customer, and what you owe them.
Agents propose. Policy decides. People govern.: approval as policy, single-use approvals and one audit record.
Built on the standards the industry already agreed: TM Forum, CAMARA, 3GPP, ETSI and Mplify, and the AI-native direction.

Go deeper. The rest of the wiki covers Exosphere Agentic in ODA terms, the entity chain across BSS and OSS, the SID-to-ontology mapping, the governance model, the two MCP surfaces and the path to your first loop. Members also get briefings for their role, cited search and proposals.

Prefer long-form reading? The white paper Approval Is a Policy, Not a Boolean sets out why approval belongs in policy, and part two shows how it is built. The library collects them with PDF briefings to download, open to everyone.

Start with one loop

Book a demo, ask for the ontology briefing, or open the library.